import numpy as np from matplotlib import rcParams from utils import Saver, TestUtils from matplotlib import rcParams rcParams['font.family'] = 'sans-serif' rcParams['font.sans-serif'] = ['Tahoma'] rcParams['font.size'] = 16 saver = Saver() test_utils = TestUtils() q_learner = saver.load_from_pickle("results/q_learner_example.pickle") lr_learner = saver.load_from_pickle("results/low_rank_learner_example.pickle") steps_q_large = saver.load_from_pickle("results/exp_1_q_learning_steps.pickle") rewards_q_large = saver.load_from_pickle( "results/exp_1_q_learning_rewards.pickle") final_mean_reward_q_large = saver.load_from_pickle( "results/exp_1_q_learning_final_reward.pickle") steps_q_small = saver.load_from_pickle("results/exp_2_q_learning_steps.pickle") rewards_q_small = saver.load_from_pickle( "results/exp_2_q_learning_rewards.pickle") final_mean_reward_q_small = saver.load_from_pickle( "results/exp_2_q_learning_final_reward.pickle") steps_lr = saver.load_from_pickle("results/exp_1_lr_learning_steps.pickle") rewards_lr = saver.load_from_pickle("results/exp_1_lr_learning_rewards.pickle") final_mean_reward_lr = saver.load_from_pickle(
from matplotlib import rcParams import matplotlib.pyplot as plt from utils import Saver import numpy as np saver = Saver() rcParams['font.family'] = 'sans-serif' rcParams['font.sans-serif'] = ['Tahoma'] rcParams['font.size'] = 16 rewards_dqn_light = saver.load_from_pickle("results/rewards_1_layer_2000_light.pck") steps_dqn_light = saver.load_from_pickle("results/steps_1_layer_2000_light.pck") rewards_dqn_large = saver.load_from_pickle("results/rewards_1_layer_2000_large.pck") steps_dqn_large = saver.load_from_pickle("results/steps_1_layer_2000_large.pck") rewards_lr = saver.load_from_pickle("results/rewards_k_2.pck") steps_lr = saver.load_from_pickle("results/steps_k_2.pck") rewards_lr_norm = saver.load_from_pickle("results/rewards_k_2_norm.pck") steps_lr_norm = saver.load_from_pickle("results/steps_k_2_norm.pck") median_rewards_dqn_light = np.median(rewards_dqn_light, axis=0) median_steps_dqn_light = np.median(steps_dqn_light, axis=0) median_rewards_dqn_large = np.median(rewards_dqn_large, axis=0) median_steps_dqn_large = np.median(steps_dqn_large, axis=0) median_reward_lr = np.median(rewards_lr, axis=0) median_steps_lr = np.median(steps_lr, axis=0)
from matplotlib import rcParams from utils import Saver, TestUtils saver = Saver() test_utils = TestUtils() rcParams['font.family'] = 'sans-serif' rcParams['font.sans-serif'] = ['Tahoma'] rcParams['font.size'] = 16 medians_q_learning = saver.load_from_pickle( "results/q_learning_medians.pickle") stds_q_learning = saver.load_from_pickle("results/q_learning_stds.pickle") frob_errors_q_learning = saver.load_from_pickle( "results/q_learning_frob_errors.pickle") colors = ['b', 'r', 'g', 'y'] epsilons = sorted([float(epsilon) for epsilon in medians_q_learning.keys()]) medians_lr_learning = saver.load_from_pickle( "results/lr_learning_medians.pickle") stds_lr_learning = saver.load_from_pickle("results/lr_learning_stds.pickle") frob_errors_lr_learning = saver.load_from_pickle( "results/lr_learning_frob_errors.pickle") test_utils.plot_smoothed_steps(medians_q=medians_q_learning, medians_lr=medians_lr_learning, epsilons=epsilons, colors=colors) test_utils.plot_sfe(epsilons=epsilons,
import json import numpy as np from utils import LowRankLearning, Saver, TestUtils, get_env parameters_file = "experiments/exp_lr_learning.json" env = get_env() saver = Saver() test_utils = TestUtils() Q_optimal = saver.load_from_pickle("results/Q_optimal.pickle") with open(parameters_file) as j: parameters = json.loads(j.read()) medians = {} standard_devs = {} frob_errors = {} for epsilon in parameters["epsilons"]: medians_temp = [] standard_devs_temp = [] frob_errors_temp = [] for i in range(parameters["n_simulations"]): lr_learner = LowRankLearning( env=env, episodes=parameters["episodes"], max_steps=parameters["max_steps"], epsilon=epsilon, gamma=parameters["gamma"], k=parameters["k"],