amoudgl

# random guessing algorithm 
# generate 10000 random configurations of the model's parameters and pick the one that achieves the best cumulative reward. 
# optimize it for weighted sum 

import gym
from gym import wrappers
import numpy as np

env = gym.make('CartPole-v0')
env = wrappers.Monitor(env, '/tmp/cartpole-random-guessing', force=True)

amoudgl

# hill climbing algorithm 
# generate a random configuration of the parameters, add small amount of noise to the parameters and evaluate the new parameter configuration
# if new configuration is better than old one, discard the old one and accept the new one
# optimize it for weighted sum 

# returns the net episode reward 
def get_episode_reward(env, observation, params):
    t = 0
    net_reward = 0
    while (t < 1000):

amoudgl

# monte carlo policy gradient algorithm 
# use neural network to decide the policy
# from observations and rewards, update the parameters of the neural networks to optimize the policy

import numpy as np
import tensorflow as tf
import gym
from gym import wrappers

# initialize constants