# parameters
n_epochs = 10000
# environment, agent
env = CatchBall(n_rows, n_cols, n_playerlength)
agent = DQNAgent(env.enable_actions, env.name, env.screen_n_rows, env.screen_n_cols, env.player_length)
# variables
win = 0
for e in range(n_epochs):
# reset
frame = 0
loss = 0.0
Q_max = 0.0
env.reset(keepPos=True)
state_t_1, reward_t, terminal = env.observe()
while not terminal:
state_t = state_t_1
# execute action in environment
action_t = agent.select_action(state_t, agent.exploration)
env.execute_action(action_t)
# observe environment
state_t_1, reward_t, terminal = env.observe()
# store experience
agent.store_experience(state_t, action_t, reward_t, state_t_1, terminal)
# parameters
n_epochs = 1000
# environment, agent
env = CatchBall()
agent = DQNAgent(env.enable_actions, env.name)
# variables
win = 0
for e in range(n_epochs):
# reset
frame = 0
loss = 0.0
Q_max = 0.0
env.reset()
state_t_1, reward_t, terminal = env.observe()
while not terminal:
state_t = state_t_1
# execute action in environment
action_t = agent.select_action(state_t, agent.exploration)
env.execute_action(action_t)
# observe environment
state_t_1, reward_t, terminal = env.observe()
# store experience
agent.store_experience(state_t, action_t, reward_t, state_t_1,
terminal)
# parameters
n_epochs = 1000
# environment, agent
env = CatchBall()
agent = DQNAgent(env.enable_actions, env.name)
# variables
win = 0
for e in range(n_epochs):
# reset
frame = 0
loss = 0.0
Q_max = 0.0
env.reset()
state_t_1, reward_t, terminal = env.observe()
while not terminal:
state_t = state_t_1
# execute action in environment
action_t = agent.select_action(state_t, agent.exploration)
env.execute_action(action_t)
# observe environment
state_t_1, reward_t, terminal = env.observe()
# store experience
agent.store_experience(state_t, action_t, reward_t, state_t_1, terminal)
