examples/RL_to_combine_search_methods.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "66f0d964",
   "metadata": {},
   "source": [
    "# Using reinforcement learning for combining search methods"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "3b6def1a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "from matplotlib.colors import ListedColormap\n",
    "\n",
    "# import loss function\n",
    "from black_it.loss_functions.msm import MethodOfMomentsLoss\n",
    "\n",
    "# import search methods\n",
    "from black_it.samplers.halton import HaltonSampler\n",
    "from black_it.samplers.best_batch import BestBatchSampler\n",
    "from black_it.samplers.random_forest import RandomForestSampler\n",
    "from black_it.samplers.xgboost import XGBoostSampler\n",
    "\n",
    "# import RL classes\n",
    "from black_it.schedulers.rl.rl_scheduler import RLScheduler\n",
    "from black_it.schedulers.rl.agents.epsilon_greedy import MABEpsilonGreedy\n",
    "from black_it.schedulers.rl.envs.mab import MABCalibrationEnv\n",
    "\n",
    "# import Calibrator class\n",
    "from black_it.calibrator import Calibrator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "5a54291c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# define a simple BH4 model as in Dyer et al. (2022)\n",
    "def BH4(theta, N, seed):\n",
    "    \"\"\"\n",
    "    Model from Brock and Hommes 1998.\n",
    "    \n",
    "    Eqs. (36-40) of Dyer et al., Black-box Bayesian inference for economic agent-based models (2022)\n",
    "    \n",
    "    Args:\n",
    "        theta: parameters\n",
    "        N: length of simulation\n",
    "        seed: random seed\n",
    "\n",
    "    Returns:\n",
    "        simulated series\n",
    "    \"\"\"\n",
    "    np.random.seed(seed=seed)\n",
    "\n",
    "    R = 1.0\n",
    "    sigma = 0.04\n",
    "    \n",
    "    beta = 120 # suggested values: 10, 120\n",
    "    g1, b1, b4 = 0., 0., 0.\n",
    "    g4 = 1.01\n",
    "    \n",
    "    g2, b2, g3, b3 = theta[0], theta[1], theta[2], theta[3]\n",
    "    # parameter set1: (g2, b2, g3, b3) = (0.9, 0.2, 0.9, −0.2)\n",
    "    x_lag3 = 0.10\n",
    "    x_lag2 = 0.10\n",
    "    x_lag1 = 0.10\n",
    "\n",
    "    x = np.zeros(N + 2)\n",
    "    n = np.full(2, 0.50)\n",
    "\n",
    "    x[0],x[1],x[2] = x_lag1, x_lag2, x_lag3\n",
    "\n",
    "    g = np.array([g1, g2, g3, g4])\n",
    "    b = np.array([b1, b2, b3, b4])\n",
    "    \n",
    "    for t in range(3, N + 1):\n",
    "        den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
    "        \n",
    "        f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R \n",
    "        \n",
    "        x[t+1] = np.random.normal(f, sigma/R)\n",
    "        \n",
    "    return np.atleast_2d(x[2:]).T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "e8c0056e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'y(t)')"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# generate synthetic target series (parameter set 1 of [Dyer et al. (2022)])\n",
    "target_params = [0.9, 0.2, 0.9, -0.2]\n",
    "target_series = BH4(theta = target_params, N= 500, seed=0)\n",
    "plt.plot(target_series)\n",
    "plt.title(\"targe time series\")\n",
    "plt.xlabel(\"t\")\n",
    "plt.ylabel(\"y(t)\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "86fdfc52",
   "metadata": {},
   "outputs": [],
   "source": [
    "# define a method-of-moments loss function\n",
    "loss = MethodOfMomentsLoss()\n",
    "\n",
    "# define some bounds and precisions on the parameters\n",
    "bounds = [[0.0, 0.0, 0.0, -1.0], \n",
    "          [1.0, 1.0, 1.0, 0.0]]\n",
    "precisions = [0.01, 0.01, 0.01, 0.01]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "b3590744",
   "metadata": {},
   "outputs": [],
   "source": [
    "# define a MAB learning agent and a MAB environment\n",
    "alpha, eps = 0.1, 0.1\n",
    "agent = MABEpsilonGreedy(alpha=alpha, eps=eps, initial_values=0.05, n_actions = 3)\n",
    "env = MABCalibrationEnv(nb_samplers=3)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "8c0a3970",
   "metadata": {},
   "outputs": [],
   "source": [
    "# define a set of samplers, and pass them to a reinforcement learning scheduler\n",
    "best_batch = BestBatchSampler(1)\n",
    "random_forest = RandomForestSampler(1)\n",
    "xg_boost = XGBoostSampler(1)\n",
    "\n",
    "rl_scheduler = RLScheduler(agent=agent, env=env, samplers = [random_forest, xg_boost, best_batch])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "b0ac6a29",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "***\n",
      "Number of free params:       4.\n",
      "Explorable param space size: 104060401.\n",
      "***\n",
      "\n",
      "Selecting 1 processes for the parallel evaluation of the model\n"
     ]
    }
   ],
   "source": [
    "# initialise a calibrator object\n",
    "cal = Calibrator(\n",
    "    scheduler=rl_scheduler,\n",
    "    loss_function=loss,\n",
    "    model=BH4,\n",
    "    parameters_bounds=bounds,\n",
    "    sim_length = 500,   # simulation length for each parameter combination\n",
    "    ensemble_size = 1,     # number of repetitions for each parameter combination\n",
    "    parameters_precision=precisions,\n",
    "    real_data=target_series,\n",
    "    random_state=0,\n",
    "    n_jobs = 1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "7701c569",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "BATCH NUMBER:   1\n",
      "PARAMS SAMPLED: 0\n",
      "\n",
      "METHOD: HaltonSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 169569.13\n",
      "---->   avg loss new params: 169569.13\n",
      "----> avg loss exist params: 169569.13\n",
      "---->         curr min loss: 169569.1271084015\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   2\n",
      "PARAMS SAMPLED: 1\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 5.51\n",
      "---->   avg loss new params: 5.51\n",
      "----> avg loss exist params: 84787.32\n",
      "---->         curr min loss: 5.511641321161405\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   3\n",
      "PARAMS SAMPLED: 2\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 4.63\n",
      "---->   avg loss new params: 4.63\n",
      "----> avg loss exist params: 56526.42\n",
      "---->         curr min loss: 4.628565393502145\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   4\n",
      "PARAMS SAMPLED: 3\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 9.2\n",
      "---->   avg loss new params: 9.2\n",
      "----> avg loss exist params: 42397.12\n",
      "---->         curr min loss: 4.628565393502145\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   5\n",
      "PARAMS SAMPLED: 4\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 3.01\n",
      "---->   avg loss new params: 3.01\n",
      "----> avg loss exist params: 33918.3\n",
      "---->         curr min loss: 3.00799724395391\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   6\n",
      "PARAMS SAMPLED: 5\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 28266.95\n",
      "---->         curr min loss: 3.00799724395391\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   7\n",
      "PARAMS SAMPLED: 6\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 6.66\n",
      "---->   avg loss new params: 6.66\n",
      "----> avg loss exist params: 24229.77\n",
      "---->         curr min loss: 3.00799724395391\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   8\n",
      "PARAMS SAMPLED: 7\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in divide\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 21202.33\n",
      "---->         curr min loss: 3.00799724395391\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   9\n",
      "PARAMS SAMPLED: 8\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.47\n",
      "---->   avg loss new params: 0.47\n",
      "----> avg loss exist params: 18846.57\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   10\n",
      "PARAMS SAMPLED: 9\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.74\n",
      "---->   avg loss new params: 0.74\n",
      "----> avg loss exist params: 16961.98\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.9s\n",
      "\n",
      "BATCH NUMBER:   11\n",
      "PARAMS SAMPLED: 10\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 13.45\n",
      "---->   avg loss new params: 13.45\n",
      "----> avg loss exist params: 15421.21\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   12\n",
      "PARAMS SAMPLED: 11\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in divide\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 14136.96\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   13\n",
      "PARAMS SAMPLED: 12\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 13050.29\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   14\n",
      "PARAMS SAMPLED: 13\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.2\n",
      "---->   avg loss new params: 2.2\n",
      "----> avg loss exist params: 12118.28\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.9s\n",
      "\n",
      "BATCH NUMBER:   15\n",
      "PARAMS SAMPLED: 14\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 11311.08\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   16\n",
      "PARAMS SAMPLED: 15\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 4.44\n",
      "---->   avg loss new params: 4.44\n",
      "----> avg loss exist params: 10604.41\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.9s\n",
      "\n",
      "BATCH NUMBER:   17\n",
      "PARAMS SAMPLED: 16\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.63\n",
      "---->   avg loss new params: 0.63\n",
      "----> avg loss exist params: 9980.66\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   18\n",
      "PARAMS SAMPLED: 17\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 9426.75\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   19\n",
      "PARAMS SAMPLED: 18\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in divide\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 8931.14\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   20\n",
      "PARAMS SAMPLED: 19\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.86\n",
      "---->   avg loss new params: 0.86\n",
      "----> avg loss exist params: 8484.63\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   21\n",
      "PARAMS SAMPLED: 20\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.75\n",
      "---->   avg loss new params: 2.75\n",
      "----> avg loss exist params: 8080.73\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   22\n",
      "PARAMS SAMPLED: 21\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in divide\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 7713.89\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   23\n",
      "PARAMS SAMPLED: 22\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 15.83\n",
      "---->   avg loss new params: 15.83\n",
      "----> avg loss exist params: 7379.19\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   24\n",
      "PARAMS SAMPLED: 23\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.5\n",
      "---->   avg loss new params: 0.5\n",
      "----> avg loss exist params: 7071.75\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   25\n",
      "PARAMS SAMPLED: 24\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.71\n",
      "---->   avg loss new params: 0.71\n",
      "----> avg loss exist params: 6788.91\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   26\n",
      "PARAMS SAMPLED: 25\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 6528.19\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   27\n",
      "PARAMS SAMPLED: 26\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.47\n",
      "---->   avg loss new params: 2.47\n",
      "----> avg loss exist params: 6286.5\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   28\n",
      "PARAMS SAMPLED: 27\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 41449.87\n",
      "---->   avg loss new params: 41449.87\n",
      "----> avg loss exist params: 7542.33\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.3s\n",
      "\n",
      "BATCH NUMBER:   29\n",
      "PARAMS SAMPLED: 28\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 5.13\n",
      "---->   avg loss new params: 5.13\n",
      "----> avg loss exist params: 7282.43\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   30\n",
      "PARAMS SAMPLED: 29\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.5\n",
      "---->   avg loss new params: 0.5\n",
      "----> avg loss exist params: 7039.7\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   31\n",
      "PARAMS SAMPLED: 30\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 6812.94\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   32\n",
      "PARAMS SAMPLED: 31\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.26\n",
      "---->   avg loss new params: 2.26\n",
      "----> avg loss exist params: 6600.11\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   33\n",
      "PARAMS SAMPLED: 32\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.94\n",
      "---->   avg loss new params: 1.94\n",
      "----> avg loss exist params: 6400.16\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   34\n",
      "PARAMS SAMPLED: 33\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 3.69\n",
      "---->   avg loss new params: 3.69\n",
      "----> avg loss exist params: 6212.03\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   35\n",
      "PARAMS SAMPLED: 34\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.66\n",
      "---->   avg loss new params: 2.66\n",
      "----> avg loss exist params: 6034.62\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   36\n",
      "PARAMS SAMPLED: 35\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.65\n",
      "---->   avg loss new params: 0.65\n",
      "----> avg loss exist params: 5867.01\n",
      "---->         curr min loss: 0.4675674394622253\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   37\n",
      "PARAMS SAMPLED: 36\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.44\n",
      "---->   avg loss new params: 0.44\n",
      "----> avg loss exist params: 5708.45\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   38\n",
      "PARAMS SAMPLED: 37\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 5558.5\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   39\n",
      "PARAMS SAMPLED: 38\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 52848.61\n",
      "---->   avg loss new params: 52848.61\n",
      "----> avg loss exist params: 6771.07\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 1.4s\n",
      "\n",
      "BATCH NUMBER:   40\n",
      "PARAMS SAMPLED: 39\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.65\n",
      "---->   avg loss new params: 1.65\n",
      "----> avg loss exist params: 6601.83\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   41\n",
      "PARAMS SAMPLED: 40\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.66\n",
      "---->   avg loss new params: 0.66\n",
      "----> avg loss exist params: 6440.83\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   42\n",
      "PARAMS SAMPLED: 41\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.06\n",
      "---->   avg loss new params: 1.06\n",
      "----> avg loss exist params: 6287.5\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 1.8s\n",
      "\n",
      "BATCH NUMBER:   43\n",
      "PARAMS SAMPLED: 42\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 7.73\n",
      "---->   avg loss new params: 7.73\n",
      "----> avg loss exist params: 6141.46\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   44\n",
      "PARAMS SAMPLED: 43\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.03\n",
      "---->   avg loss new params: 2.03\n",
      "----> avg loss exist params: 6001.93\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   45\n",
      "PARAMS SAMPLED: 44\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 5868.78\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   46\n",
      "PARAMS SAMPLED: 45\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.61\n",
      "---->   avg loss new params: 0.61\n",
      "----> avg loss exist params: 5741.21\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   47\n",
      "PARAMS SAMPLED: 46\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 14.09\n",
      "---->   avg loss new params: 14.09\n",
      "----> avg loss exist params: 5619.36\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   48\n",
      "PARAMS SAMPLED: 47\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.47\n",
      "---->   avg loss new params: 0.47\n",
      "----> avg loss exist params: 5502.3\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   49\n",
      "PARAMS SAMPLED: 48\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 5390.21\n",
      "---->         curr min loss: 0.4434231579477171\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   50\n",
      "PARAMS SAMPLED: 49\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.42\n",
      "---->   avg loss new params: 0.42\n",
      "----> avg loss exist params: 5282.42\n",
      "---->         curr min loss: 0.42233888289795174\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   51\n",
      "PARAMS SAMPLED: 50\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.6\n",
      "---->   avg loss new params: 0.6\n",
      "----> avg loss exist params: 5178.85\n",
      "---->         curr min loss: 0.42233888289795174\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   52\n",
      "PARAMS SAMPLED: 51\n",
      "\n",
      "METHOD: BestBatchSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.39\n",
      "---->   avg loss new params: 0.39\n",
      "----> avg loss exist params: 5079.27\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   53\n",
      "PARAMS SAMPLED: 52\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 6.49\n",
      "---->   avg loss new params: 6.49\n",
      "----> avg loss exist params: 4983.55\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   54\n",
      "PARAMS SAMPLED: 53\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.43\n",
      "---->   avg loss new params: 0.43\n",
      "----> avg loss exist params: 4891.27\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   55\n",
      "PARAMS SAMPLED: 54\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.97\n",
      "---->   avg loss new params: 0.97\n",
      "----> avg loss exist params: 4802.36\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   56\n",
      "PARAMS SAMPLED: 55\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 3.03\n",
      "---->   avg loss new params: 3.03\n",
      "----> avg loss exist params: 4716.66\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   57\n",
      "PARAMS SAMPLED: 56\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.77\n",
      "---->   avg loss new params: 1.77\n",
      "----> avg loss exist params: 4633.94\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   58\n",
      "PARAMS SAMPLED: 57\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.43\n",
      "---->   avg loss new params: 0.43\n",
      "----> avg loss exist params: 4554.05\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   59\n",
      "PARAMS SAMPLED: 58\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.86\n",
      "---->   avg loss new params: 0.86\n",
      "----> avg loss exist params: 4476.88\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   60\n",
      "PARAMS SAMPLED: 59\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.39\n",
      "---->   avg loss new params: 1.39\n",
      "----> avg loss exist params: 4402.29\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   61\n",
      "PARAMS SAMPLED: 60\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.1s\n",
      "---->   min loss new params: 3.29\n",
      "---->   avg loss new params: 3.29\n",
      "----> avg loss exist params: 4330.17\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.8s\n",
      "\n",
      "BATCH NUMBER:   62\n",
      "PARAMS SAMPLED: 61\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.48\n",
      "---->   avg loss new params: 0.48\n",
      "----> avg loss exist params: 4260.34\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   63\n",
      "PARAMS SAMPLED: 62\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.26\n",
      "---->   avg loss new params: 2.26\n",
      "----> avg loss exist params: 4192.75\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   64\n",
      "PARAMS SAMPLED: 63\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 15.4\n",
      "---->   avg loss new params: 15.4\n",
      "----> avg loss exist params: 4127.48\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   65\n",
      "PARAMS SAMPLED: 64\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.4\n",
      "---->   avg loss new params: 0.4\n",
      "----> avg loss exist params: 4063.99\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   66\n",
      "PARAMS SAMPLED: 65\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.24\n",
      "---->   avg loss new params: 1.24\n",
      "----> avg loss exist params: 4002.43\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   67\n",
      "PARAMS SAMPLED: 66\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.69\n",
      "---->   avg loss new params: 0.69\n",
      "----> avg loss exist params: 3942.7\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   68\n",
      "PARAMS SAMPLED: 67\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.65\n",
      "---->   avg loss new params: 1.65\n",
      "----> avg loss exist params: 3884.74\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   69\n",
      "PARAMS SAMPLED: 68\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.3\n",
      "---->   avg loss new params: 2.3\n",
      "----> avg loss exist params: 3828.48\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.7s\n",
      "\n",
      "BATCH NUMBER:   70\n",
      "PARAMS SAMPLED: 69\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 3773.93\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   71\n",
      "PARAMS SAMPLED: 70\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.21\n",
      "---->   avg loss new params: 2.21\n",
      "----> avg loss exist params: 3720.81\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   72\n",
      "PARAMS SAMPLED: 71\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 6.9\n",
      "---->   avg loss new params: 6.9\n",
      "----> avg loss exist params: 3669.23\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.5s\n",
      "\n",
      "BATCH NUMBER:   73\n",
      "PARAMS SAMPLED: 72\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 3619.1\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   74\n",
      "PARAMS SAMPLED: 73\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.47\n",
      "---->   avg loss new params: 0.47\n",
      "----> avg loss exist params: 3570.2\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   75\n",
      "PARAMS SAMPLED: 74\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.44\n",
      "---->   avg loss new params: 0.44\n",
      "----> avg loss exist params: 3522.61\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 1.3s\n",
      "\n",
      "BATCH NUMBER:   76\n",
      "PARAMS SAMPLED: 75\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 3476.39\n",
      "---->         curr min loss: 0.3885017552896394\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   77\n",
      "PARAMS SAMPLED: 76\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.1s\n",
      "---->   min loss new params: 0.36\n",
      "---->   avg loss new params: 0.36\n",
      "----> avg loss exist params: 3431.25\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   78\n",
      "PARAMS SAMPLED: 77\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.41\n",
      "---->   avg loss new params: 0.41\n",
      "----> avg loss exist params: 3387.26\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   79\n",
      "PARAMS SAMPLED: 78\n",
      "\n",
      "METHOD: BestBatchSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.1s\n",
      "---->   min loss new params: 0.4\n",
      "---->   avg loss new params: 0.4\n",
      "----> avg loss exist params: 3344.39\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   80\n",
      "PARAMS SAMPLED: 79\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.97\n",
      "---->   avg loss new params: 1.97\n",
      "----> avg loss exist params: 3302.61\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   81\n",
      "PARAMS SAMPLED: 80\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.8\n",
      "---->   avg loss new params: 1.8\n",
      "----> avg loss exist params: 3261.86\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   82\n",
      "PARAMS SAMPLED: 81\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.81\n",
      "---->   avg loss new params: 0.81\n",
      "----> avg loss exist params: 3222.09\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   83\n",
      "PARAMS SAMPLED: 82\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.43\n",
      "---->   avg loss new params: 0.43\n",
      "----> avg loss exist params: 3183.28\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   84\n",
      "PARAMS SAMPLED: 83\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.42\n",
      "---->   avg loss new params: 0.42\n",
      "----> avg loss exist params: 3145.38\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   85\n",
      "PARAMS SAMPLED: 84\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 14.71\n",
      "---->   avg loss new params: 14.71\n",
      "----> avg loss exist params: 3108.55\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   86\n",
      "PARAMS SAMPLED: 85\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.49\n",
      "---->   avg loss new params: 1.49\n",
      "----> avg loss exist params: 3072.42\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   87\n",
      "PARAMS SAMPLED: 86\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 3.13\n",
      "---->   avg loss new params: 3.13\n",
      "----> avg loss exist params: 3037.14\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   88\n",
      "PARAMS SAMPLED: 87\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.96\n",
      "---->   avg loss new params: 0.96\n",
      "----> avg loss exist params: 3002.64\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   89\n",
      "PARAMS SAMPLED: 88\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.55\n",
      "---->   avg loss new params: 2.55\n",
      "----> avg loss exist params: 2968.93\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   90\n",
      "PARAMS SAMPLED: 89\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.41\n",
      "---->   avg loss new params: 0.41\n",
      "----> avg loss exist params: 2935.95\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   91\n",
      "PARAMS SAMPLED: 90\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 5.99\n",
      "---->   avg loss new params: 5.99\n",
      "----> avg loss exist params: 2903.75\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   92\n",
      "PARAMS SAMPLED: 91\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.53\n",
      "---->   avg loss new params: 0.53\n",
      "----> avg loss exist params: 2872.2\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   93\n",
      "PARAMS SAMPLED: 92\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in divide\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2841.42\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   94\n",
      "PARAMS SAMPLED: 93\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2811.3\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   95\n",
      "PARAMS SAMPLED: 94\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2781.82\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   96\n",
      "PARAMS SAMPLED: 95\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.57\n",
      "---->   avg loss new params: 0.57\n",
      "----> avg loss exist params: 2752.85\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   97\n",
      "PARAMS SAMPLED: 96\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.34\n",
      "---->   avg loss new params: 2.34\n",
      "----> avg loss exist params: 2724.49\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 1.1s\n",
      "\n",
      "BATCH NUMBER:   98\n",
      "PARAMS SAMPLED: 97\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2696.8\n",
      "---->         curr min loss: 0.3559064883487337\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   99\n",
      "PARAMS SAMPLED: 98\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.32\n",
      "---->   avg loss new params: 0.32\n",
      "----> avg loss exist params: 2669.56\n",
      "---->         curr min loss: 0.32396118971466287\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   100\n",
      "PARAMS SAMPLED: 99\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.02\n",
      "---->   avg loss new params: 2.02\n",
      "----> avg loss exist params: 2642.88\n",
      "---->         curr min loss: 0.32396118971466287\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   101\n",
      "PARAMS SAMPLED: 100\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.31\n",
      "---->   avg loss new params: 0.31\n",
      "----> avg loss exist params: 2616.72\n",
      "---->         curr min loss: 0.31228142098336037\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   102\n",
      "PARAMS SAMPLED: 101\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.36\n",
      "---->   avg loss new params: 0.36\n",
      "----> avg loss exist params: 2591.07\n",
      "---->         curr min loss: 0.31228142098336037\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   103\n",
      "PARAMS SAMPLED: 102\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.26\n",
      "---->   avg loss new params: 0.26\n",
      "----> avg loss exist params: 2565.91\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   104\n",
      "PARAMS SAMPLED: 103\n",
      "\n",
      "METHOD: BestBatchSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.29\n",
      "---->   avg loss new params: 0.29\n",
      "----> avg loss exist params: 2541.25\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   105\n",
      "PARAMS SAMPLED: 104\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.55\n",
      "---->   avg loss new params: 1.55\n",
      "----> avg loss exist params: 2517.06\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   106\n",
      "PARAMS SAMPLED: 105\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.36\n",
      "---->   avg loss new params: 0.36\n",
      "----> avg loss exist params: 2493.32\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   107\n",
      "PARAMS SAMPLED: 106\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.33\n",
      "---->   avg loss new params: 0.33\n",
      "----> avg loss exist params: 2470.02\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   108\n",
      "PARAMS SAMPLED: 107\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.34\n",
      "---->   avg loss new params: 0.34\n",
      "----> avg loss exist params: 2447.15\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   109\n",
      "PARAMS SAMPLED: 108\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.34\n",
      "---->   avg loss new params: 0.34\n",
      "----> avg loss exist params: 2424.7\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   110\n",
      "PARAMS SAMPLED: 109\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.56\n",
      "---->   avg loss new params: 0.56\n",
      "----> avg loss exist params: 2402.66\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   111\n",
      "PARAMS SAMPLED: 110\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in divide\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2381.11\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   112\n",
      "PARAMS SAMPLED: 111\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.29\n",
      "---->   avg loss new params: 0.29\n",
      "----> avg loss exist params: 2359.85\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   113\n",
      "PARAMS SAMPLED: 112\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.31\n",
      "---->   avg loss new params: 0.31\n",
      "----> avg loss exist params: 2338.97\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   114\n",
      "PARAMS SAMPLED: 113\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.4\n",
      "---->   avg loss new params: 0.4\n",
      "----> avg loss exist params: 2318.46\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   115\n",
      "PARAMS SAMPLED: 114\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.35\n",
      "---->   avg loss new params: 0.35\n",
      "----> avg loss exist params: 2298.3\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   116\n",
      "PARAMS SAMPLED: 115\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2278.58\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   117\n",
      "PARAMS SAMPLED: 116\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.11\n",
      "---->   avg loss new params: 2.11\n",
      "----> avg loss exist params: 2259.12\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   118\n",
      "PARAMS SAMPLED: 117\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.4\n",
      "---->   avg loss new params: 0.4\n",
      "----> avg loss exist params: 2239.98\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   119\n",
      "PARAMS SAMPLED: 118\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.36\n",
      "---->   avg loss new params: 0.36\n",
      "----> avg loss exist params: 2221.16\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   120\n",
      "PARAMS SAMPLED: 119\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.34\n",
      "---->   avg loss new params: 0.34\n",
      "----> avg loss exist params: 2202.65\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   121\n",
      "PARAMS SAMPLED: 120\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 2184.53\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   122\n",
      "PARAMS SAMPLED: 121\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.45\n",
      "---->   avg loss new params: 0.45\n",
      "----> avg loss exist params: 2166.63\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   123\n",
      "PARAMS SAMPLED: 122\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.3\n",
      "---->   avg loss new params: 2.3\n",
      "----> avg loss exist params: 2149.03\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   124\n",
      "PARAMS SAMPLED: 123\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.27\n",
      "---->   avg loss new params: 1.27\n",
      "----> avg loss exist params: 2131.71\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   125\n",
      "PARAMS SAMPLED: 124\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.2\n",
      "---->   avg loss new params: 2.2\n",
      "----> avg loss exist params: 2114.68\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   126\n",
      "PARAMS SAMPLED: 125\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.58\n",
      "---->   avg loss new params: 1.58\n",
      "----> avg loss exist params: 2097.91\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   127\n",
      "PARAMS SAMPLED: 126\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.16\n",
      "---->   avg loss new params: 2.16\n",
      "----> avg loss exist params: 2081.4\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   128\n",
      "PARAMS SAMPLED: 127\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.67\n",
      "---->   avg loss new params: 0.67\n",
      "----> avg loss exist params: 2065.15\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.0s\n",
      "\n",
      "BATCH NUMBER:   129\n",
      "PARAMS SAMPLED: 128\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.43\n",
      "---->   avg loss new params: 1.43\n",
      "----> avg loss exist params: 2049.15\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   130\n",
      "PARAMS SAMPLED: 129\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.45\n",
      "---->   avg loss new params: 0.45\n",
      "----> avg loss exist params: 2033.39\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   131\n",
      "PARAMS SAMPLED: 130\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.89\n",
      "---->   avg loss new params: 2.89\n",
      "----> avg loss exist params: 2017.89\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   132\n",
      "PARAMS SAMPLED: 131\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 6.65\n",
      "---->   avg loss new params: 6.65\n",
      "----> avg loss exist params: 2002.65\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   133\n",
      "PARAMS SAMPLED: 132\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.05\n",
      "---->   avg loss new params: 2.05\n",
      "----> avg loss exist params: 1987.61\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   134\n",
      "PARAMS SAMPLED: 133\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.34\n",
      "---->   avg loss new params: 0.34\n",
      "----> avg loss exist params: 1972.78\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.3s\n",
      "\n",
      "BATCH NUMBER:   135\n",
      "PARAMS SAMPLED: 134\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 5.66\n",
      "---->   avg loss new params: 5.66\n",
      "----> avg loss exist params: 1958.21\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   136\n",
      "PARAMS SAMPLED: 135\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.49\n",
      "---->   avg loss new params: 0.49\n",
      "----> avg loss exist params: 1943.82\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   137\n",
      "PARAMS SAMPLED: 136\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 2.76\n",
      "---->   avg loss new params: 2.76\n",
      "----> avg loss exist params: 1929.65\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   138\n",
      "PARAMS SAMPLED: 137\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 10.24\n",
      "---->   avg loss new params: 10.24\n",
      "----> avg loss exist params: 1915.74\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   139\n",
      "PARAMS SAMPLED: 138\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.32\n",
      "---->   avg loss new params: 0.32\n",
      "----> avg loss exist params: 1901.96\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   140\n",
      "PARAMS SAMPLED: 139\n",
      "\n",
      "METHOD: RandomForestSampler\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:40: RuntimeWarning: overflow encountered in exp\n",
      "  den = np.sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])))\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: overflow encountered in exp\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n",
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/1649505445.py:42: RuntimeWarning: invalid value encountered in multiply\n",
      "  f = sum(np.exp(beta*(x[t] - R*x[t-1]) * (g*x[t-2] + b - R*x[t-1])) * (g*x[t] + b) / den) / R\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.42\n",
      "---->   avg loss new params: 0.42\n",
      "----> avg loss exist params: 1888.38\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.2s\n",
      "\n",
      "BATCH NUMBER:   141\n",
      "PARAMS SAMPLED: 140\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 13.59\n",
      "---->   avg loss new params: 13.59\n",
      "----> avg loss exist params: 1875.08\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.1s\n",
      "\n",
      "BATCH NUMBER:   142\n",
      "PARAMS SAMPLED: 141\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.28\n",
      "---->   avg loss new params: 0.28\n",
      "----> avg loss exist params: 1861.88\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   143\n",
      "PARAMS SAMPLED: 142\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.94\n",
      "---->   avg loss new params: 0.94\n",
      "----> avg loss exist params: 1848.86\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.4s\n",
      "\n",
      "BATCH NUMBER:   144\n",
      "PARAMS SAMPLED: 143\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.56\n",
      "---->   avg loss new params: 1.56\n",
      "----> avg loss exist params: 1836.03\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   145\n",
      "PARAMS SAMPLED: 144\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.32\n",
      "---->   avg loss new params: 0.32\n",
      "----> avg loss exist params: 1823.37\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   146\n",
      "PARAMS SAMPLED: 145\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.78\n",
      "---->   avg loss new params: 0.78\n",
      "----> avg loss exist params: 1810.89\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.3s\n",
      "\n",
      "BATCH NUMBER:   147\n",
      "PARAMS SAMPLED: 146\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 6.02\n",
      "---->   avg loss new params: 6.02\n",
      "----> avg loss exist params: 1798.61\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   148\n",
      "PARAMS SAMPLED: 147\n",
      "\n",
      "METHOD: BestBatchSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 1.42\n",
      "---->   avg loss new params: 1.42\n",
      "----> avg loss exist params: 1786.47\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.0s\n",
      "\n",
      "BATCH NUMBER:   149\n",
      "PARAMS SAMPLED: 148\n",
      "\n",
      "METHOD: RandomForestSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 0.61\n",
      "---->   avg loss new params: 0.61\n",
      "----> avg loss exist params: 1774.48\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 1.5s\n",
      "\n",
      "BATCH NUMBER:   150\n",
      "PARAMS SAMPLED: 149\n",
      "\n",
      "METHOD: XGBoostSampler\n",
      "----> sim exec elapsed time: 0.0s\n",
      "---->   min loss new params: 6.98\n",
      "---->   avg loss new params: 6.98\n",
      "----> avg loss exist params: 1762.7\n",
      "---->         curr min loss: 0.2649392396747757\n",
      "====>    total elapsed time: 0.1s\n"
     ]
    }
   ],
   "source": [
    "# calibrate for a number of epochs\n",
    "params, losses = cal.calibrate(150)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "4e6529f6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.9, 0.2, 0.9, -0.2]\n",
      "[ 0.63  0.41  0.8  -0.4 ]\n"
     ]
    }
   ],
   "source": [
    "# confront target patrameters with obtained parameters\n",
    "print(target_params)\n",
    "print(params[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "54aab78f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'RandomForestSampler': 0,\n",
       " 'XGBoostSampler': 1,\n",
       " 'BestBatchSampler': 2,\n",
       " 'HaltonSampler': 3}"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# integers for each method used\n",
    "cal.samplers_id_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "4de2a549",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 0, 0, 1, 2, 0, 1, 2,\n",
       "       0, 0, 1, 2, 1, 2, 2, 2, 1, 2, 2, 1, 2, 2, 0, 2, 0, 2, 0, 1, 2, 0,\n",
       "       1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2, 2, 2, 2, 2, 2, 2, 0, 1, 2, 0, 1,\n",
       "       2, 0, 1, 2, 0, 1, 1, 2, 0, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,\n",
       "       0, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 0, 2, 0, 1, 2, 0, 1, 2, 0, 1,\n",
       "       2, 0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 1])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the action selected by the RL scheduler\n",
    "cal.method_samp "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "bef3ab30",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/t5/yx84b1sd52lblk0kjl01g2y00000gn/T/ipykernel_55171/707759767.py:17: MatplotlibDeprecationWarning: Unable to determine Axes to steal space for Colorbar. Using gca(), but will raise in the future. Either provide the *cax* argument to use as the Axes for the Colorbar, provide the *ax* argument to steal space from it, or add *mappable* to an Axes.\n",
      "  cbar = plt.colorbar(matplotlib.cm.ScalarMappable(norm=norm, cmap=cmap2),\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x180 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "actions = cal.method_samp;\n",
    "actions = np.array(np.tile(actions[1:], (20,1)));\n",
    "\n",
    "cmap2 = matplotlib.cm.viridis\n",
    "cmap2_r = ListedColormap(cmap2.colors[::-1])\n",
    "bounds = [0, 1, 2, 3]\n",
    "norm = matplotlib.colors.BoundaryNorm(bounds, cmap2.N)\n",
    "\n",
    "plt.figure(figsize = (8, 1.8))\n",
    "plt.imshow(actions, aspect='auto', cmap=cmap2)\n",
    "\n",
    "plt.xlabel('epoch number')\n",
    "\n",
    "plt.yticks([])\n",
    "plt.title(f'eps={eps}, alpha={alpha}')\n",
    "\n",
    "cbar = plt.colorbar(matplotlib.cm.ScalarMappable(norm=norm, cmap=cmap2),\n",
    "              orientation='vertical', ticks=[0.5, 1.5, 2.5])\n",
    "\n",
    "cbar.ax.set_yticklabels(['RF', 'XB', 'BB'])\n",
    "\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "e060b692",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 1.0)"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# the convergence of the loss and the reward signals\n",
    "\n",
    "cum_min = np.minimum.accumulate(cal.losses_samp)\n",
    "rewards = -((np.roll(cum_min, -1) - cum_min)/cum_min)[:-1]\n",
    "\n",
    "fig, ax1 = plt.subplots()\n",
    "\n",
    "ax2 = ax1.twinx()\n",
    "lns1 = ax1.plot(cum_min[1:], color=\"tab:blue\", label=\"loss\")\n",
    "lns2 = ax2.plot(rewards, color=\"orange\", label=\"reward\")\n",
    "\n",
    "lns = lns1+lns2\n",
    "labs = [l.get_label() for l in lns]\n",
    "ax1.legend(lns, labs, loc=0)\n",
    "\n",
    "ax1.set_xlabel('Epoch')\n",
    "ax1.set_ylabel('Loss')\n",
    "ax2.set_ylabel('Reward')\n",
    "\n",
    "ax2.set_ylim(0.0, 1.0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2efb0c81",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ba9be9a5",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}