def main():
import argparse
parser = argparse.ArgumentParser(description='Train it :3')
parser.add_argument('--ckpt-path', help='relative path to checkpoint to saved model', default=None)
args = parser.parse_args()
# Leaving everything in a single function for easy later CLI argument parsing
experiment_params = dict(
exp_name='32-SimpleModel-gainx10-longrun',
# env_name='Pendulum-v0',
# env_name='DroneZero-v0',
# env_name='PendrogoneFlagrun-v0',
env_name='WindTurbine-v2',
num_iterations=600,
sample_horizon=8192,
# Learning hyperparameters
epochs=10, batch_size=128, learning_rate=1e-4,
# GAE params
gamma=0.95, lam=0.95,
# PPO specific hyperparameter, not gonna change this :v
epsilon=0.2,
ckpt_path=args.ckpt_path
)
exp_dir = experiment(n_experiments=4, **experiment_params)
data = plotter.get_datasets(exp_dir)
plotter.plot_data(data, os.path.join(exp_dir, 'plot4this.png'))
def main():
# Leaving everything in a single function for easy later CLI argument parsing
experiment_params = dict(
exp_name='06-exponential',
# env_name='Pendulum-v0',
# env_name='DroneZero-v0',
env_name='PendrogoneZero-v0',
num_iterations=350,
sample_horizon=2048,
# Learning hyperparameters
epochs=10, batch_size=64, learning_rate=1e-4,
# GAE params
gamma=0.99, lam=0.95,
# PPO specific hyperparameter, not gonna change this :v
epsilon=0.2,
)
exp_dir = experiment(n_experiments=4, **experiment_params)
data = plotter.get_datasets(exp_dir)
plotter.plot_data(data, os.path.join(exp_dir, 'plot4this.png'))
import numpy as np import matplotlib.pyplot as plt from sklearn.linear_model import LogisticRegression from exam import hours_studied, passed_exam from plotter import plot_data # Create logistic regression model model = LogisticRegression() model.fit(hours_studied,passed_exam) # Plug sample data into fitted model sample_x = np.linspace(-16.65, 33.35, 300).reshape(-1,1) probability = model.predict_proba(sample_x)[:,1] # Function to plot exam data and logistic regression curve plot_data(model) # Show the plot plt.show() # Lowest and highest probabilities lowest = 0 highest = 1 ''' LOGISTIC REGRESSION Log-Odds In Linear Regression we multiply the coefficients of our features by their respective feature values and add the intercept, resulting in our prediction, which can range from -∞ to +∞. In Logistic Regression, we make the same multiplication of feature coefficients and feature values and add the intercept, but instead of the prediction, we get what is called the log-odds.
if new_path_distance < old_path_distance:
path = new_path
if new_path_distance < min_distance:
min_distance = new_path_distance
elif math.exp(-(new_path_distance - old_path_distance) /
temperature) > rand.uniform(0, 1):
path = new_path
temperatures.append(temperature)
temperature *= decay_rate
distances_plot_data = (copy.copy(iters), dists)
temperatures_plot_data = (iters, temperatures)
return first_path, best_path, distances_plot_data, temperatures_plot_data
n = 200
iterations = 500000
temperature = 1000
decay_rate = 0.99995
swap_type = "consecutive"
low = 0
high = 1
distribution = "uniform"
first_path, best_path, distances_plot_data, temperatures_plot_data = travelling_salesman_problem(
n, iterations, temperature, decay_rate, swap_type, low, high, distribution)
plotter.plot_data(first_path, best_path, distances_plot_data,
temperatures_plot_data)