Method sample_action

rl3/es_mujoco.py:63–68

(self, x)

    return np.tanh(Z.dot(self.W2) + self.b2) * action_max

  def sample_action(self, x):
    # assume input is a single state of size (D,)
    # first make it (N, D) to fit ML conventions
    X = np.atleast_2d(x)
    Y = self.forward(X)
    return Y[0] # the first row

  def get_params(self):
    # return a flat array of parameters