76b6bcace9245a57f439d39deac653e2384a645b,PyPi/utils/dataset.py,,max_QA,#Any#Any#Any#Any#,113

Before Change

    // store Q-value and action for each state
    r_q, r_a = np.zeros(n_states), np.zeros((n_states, action_dim), dtype=int)
    for idx in xrange(n_states):
        r_q[idx] = Q[idx, amax[idx]]
        r_a[idx] = discrete_actions[amax[idx]]

    return r_q, r_a


def state_action(state, action):

After Change

        predictions = approximator.predict(samples)
        Q[:, action] = predictions * (1 - absorbing)

    max_q = np.max(Q, axis=1)
    if Q.shape[0] > 1:
        max_a = np.argmax(Q, axis=1)
    else:
        max_a = [np.random.choice(np.argwhere(Q[0] == max_q).ravel())]

    return max_q, discrete_actions[max_a]


def state_action(state, action):

Instances