Method forward

rl3/es_mujoco.py:59–61

(self, X)

    self.b2 = np.zeros(K)

  def forward(self, X):
    Z = self.f(X.dot(self.W1) + self.b1)
    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,)