def main():
list_directories = glob.glob(
path + '*/') # get a list of all directoies for different step sizes
# Loop for each lambda and plot it
for each_lam_directory in list_directories:
print(each_lam_directory)
_, lam, _ = get_method_lam_alpha(each_lam_directory)
result = get_mean_and_SEM_all_stepSizes(each_lam_directory, curve,
performance_measure)
# Sort the result according to the step size
result.sort(key=lambda x: x[0])
alphas = [each_point[0] for each_point in result]
averages = [each_point[1] for each_point in result]
errors = [each_point[2] for each_point in result]
# # Plot the curve for current lambda
plt.errorbar(alphas,
averages,
yerr=errors,
fmt='-o',
label=r'$\lambda$' + '= ' + str(lam))
plt.title(method + '\n' + curve_verbose, fontsize=15)
plt.xscale('log')
plt.xlabel(r'$\alpha$', fontsize=20)
plt.ylabel('Steps per episode', rotation=0)
plt.legend(loc='best')
plt.show()
from matplotlib import pyplot as plt
import argparse
import numpy as np
from plot_util import get_method_lam_alpha, COMMON_PATH, get_all_files
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--directory',
default='TD/Data/0.0/2-8/',
type=str,
help='directory that contains files of different runs')
args = parser.parse_args()
directory = COMMON_PATH + args.directory # Turn directory into an absolute path
method, lam, alpha = get_method_lam_alpha(
directory) # Extract method, lambda and step size from the directory path
# Read data from a directory and plot the results
def main():
# measures is a list of lists, where each element (each list)
# is the measures of a trial
measures = get_all_files(directory)
# Plot the graph
mean = np.mean(measures, axis=0)
plt.plot(np.arange(mean.size),
mean,
label=r'$\lambda$' + '=' + str(lam) + ', alpha=' + str(alpha))
plt.xlabel('Episodes')
plt.ylabel('Step per Episode\nAveraged over ' + str(len(measures)) +