def main(args): os.environ["CUDA_VISIBLE_DEVICES"] = "" # maybe create dir if not os.path.isdir(SAVE_DIR): os.makedirs(SAVE_DIR) # load results if already exist if os.path.isfile(SAVE_PATH): results = log_utils.read_pickle(SAVE_PATH) else: results = {} for num_experience in NUM_EXPERIENCE_LIST: for split_threshold in SPLIT_THRESHOLD_LIST: for min_conf in CONF_THRESHOLD_LIST: for run_idx in range(NUM_RUNS): key = (num_experience, split_threshold, min_conf, run_idx) # skip if this setting is already in results if key in results: continue accuracy = run(num_experience, split_threshold, min_conf) results[key] = accuracy # save results after each run log_utils.write_pickle(SAVE_PATH, results)
def main(args): os.environ["CUDA_VISIBLE_DEVICES"] = "" # maybe create dir if not os.path.isdir(SAVE_DIR): os.makedirs(SAVE_DIR) # load results if already exist if os.path.isfile(SAVE_PATH): results = log_utils.read_pickle(SAVE_PATH) else: results = {} for resolution in RESOLUTION_LIST: for run_idx in range(NUM_RUNS): key = (resolution, run_idx) # skip if this setting is already in results if key in results: continue accuracy = run(resolution) results[key] = accuracy # save results after each run log_utils.write_pickle(SAVE_PATH, results)
def main(args): os.environ["CUDA_VISIBLE_DEVICES"] = "" NUM_RUNS = 50 NUM_EXPERIENCE = 2000 SPLIT_THRESHOLD_LIST = [50, 100, 200, 500] SAVE_DIR = "results/homo_g/balanced_mlp" SAVE_FILE = "continuous_3_evaluation_3.pickle" SAVE_PATH = os.path.join(SAVE_DIR, SAVE_FILE) # maybe create dir if not os.path.isdir(SAVE_DIR): os.makedirs(SAVE_DIR) # load results if already exist if os.path.isfile(SAVE_PATH): results = log_utils.read_pickle(SAVE_PATH) else: results = {} for t1 in SPLIT_THRESHOLD_LIST: for t2 in SPLIT_THRESHOLD_LIST: for run_idx in range(NUM_RUNS): key = (t1, t2, run_idx) # skip if this setting is already in results if key in results: continue accuracy = run(NUM_EXPERIENCE, t1, t2) results[key] = accuracy # save results after each run log_utils.write_pickle(SAVE_PATH, results)
def main(args): os.environ["CUDA_VISIBLE_DEVICES"] = "" NUM_RUNS = 200 NUM_EXPERIENCE_LIST = [200, 500, 1000] SPLIT_THRESHOLD_LIST = [50, 100, 200] SAVE_DIR = "results/homo_g/balanced_mlp" SAVE_FILE = "experiment_1_thresholds.pickle" SAVE_PATH = os.path.join(SAVE_DIR, SAVE_FILE) # maybe create dir if not os.path.isdir(SAVE_DIR): os.makedirs(SAVE_DIR) # load results if already exist if os.path.isfile(SAVE_PATH): results = log_utils.read_pickle(SAVE_PATH) else: results = {} for num_experience in NUM_EXPERIENCE_LIST: for split_threshold in SPLIT_THRESHOLD_LIST: for run_idx in range(NUM_RUNS): key = (num_experience, split_threshold, run_idx) # skip if this setting is already in results if key in results: continue accuracy = run(num_experience, split_threshold) results[key] = accuracy # save results after each run log_utils.write_pickle(SAVE_PATH, results)
import os import numpy as np import log_utils import seaborn as sns import matplotlib.pyplot as plt LOAD_DIR = "results/homo_g_conf/balanced_mlp" LOAD_FILE = "experiment_1_thresholds_12th_percentile.pickle" LOAD_PATH = os.path.join(LOAD_DIR, LOAD_FILE) NUM_RUNS = 200 NUM_EXPERIENCE_LIST = [200, 500, 1000] SPLIT_THRESHOLD_LIST = [10, 20, 50] CONF_THRESHOLD_LIST = [0.0, 0.5, 0.7, 0.8, 0.85, 0.9] results = log_utils.read_pickle(LOAD_PATH) results_array = np.zeros((len(NUM_EXPERIENCE_LIST), len(SPLIT_THRESHOLD_LIST), len(CONF_THRESHOLD_LIST))) for i, num_experience in enumerate(NUM_EXPERIENCE_LIST): for j, split_threshold in enumerate(SPLIT_THRESHOLD_LIST): for k, min_conf in enumerate(CONF_THRESHOLD_LIST): accuracies = [] for run_idx in range(NUM_RUNS): key = (num_experience, split_threshold, min_conf, run_idx)
import matplotlib.pyplot as plt LOAD_DIR_1 = "results/homo_g_conf/balanced_mlp" LOAD_FILE_1 = "experiment_1_thresholds_25th_percentile_sort_blocks.pickle" LOAD_PATH_1 = os.path.join(LOAD_DIR_1, LOAD_FILE_1) LOAD_DIR_2 = "results/homo_g_conf/balanced_mlp" LOAD_FILE_2 = "experiment_1_thresholds_25th_percentile.pickle" LOAD_PATH_2 = os.path.join(LOAD_DIR_2, LOAD_FILE_2) NUM_RUNS = 200 NUM_EXPERIENCE_LIST = [200, 500, 1000] SPLIT_THRESHOLD_LIST = [10, 20, 50] CONF_THRESHOLD_LIST = [0.0, 0.5, 0.7, 0.8, 0.85, 0.9] results_1 = log_utils.read_pickle(LOAD_PATH_1) results_2 = log_utils.read_pickle(LOAD_PATH_2) results_array = np.zeros((len(NUM_EXPERIENCE_LIST), len(SPLIT_THRESHOLD_LIST), len(CONF_THRESHOLD_LIST))) for i, num_experience in enumerate(NUM_EXPERIENCE_LIST): for j, split_threshold in enumerate(SPLIT_THRESHOLD_LIST): for k, min_conf in enumerate(CONF_THRESHOLD_LIST): accuracies_1 = [] accuracies_2 = [] for run_idx in range(NUM_RUNS):