Function epsilon_greedy

rl/approx_control.py:23–31

(model, s, eps=0.1)

def epsilon_greedy(model, s, eps=0.1):
  # we'll use epsilon-soft to ensure all states are visited
  # what happens if you don't do this? i.e. eps=0
  p = np.random.random()
  if p < (1 - eps):
    values = model.predict_all_actions(s)
    return ALL_POSSIBLE_ACTIONS[np.argmax(values)]
  else:
    return np.random.choice(ALL_POSSIBLE_ACTIONS)


def one_hot(k):