Method entropy

rl2/mountaincar/pg_theano.py:109–110

(var)

      return -0.5*(k1 + k2)

    def entropy(var):
      return 0.5*T.log(2*np.pi*np.e*var)

    log_probs = log_pdf(actions, mean, var)
    cost = -T.sum(advantages * log_probs + 0.1*entropy(var))