Merge pull request #128 from kengz/policy-gradient

Solution: Acrobot-v1 with ActorCritic

Author: kengz

rl/spec/classic_experiment_specs.json

@@ -733,9 +733,9 @@
     "Policy": "SoftmaxPolicy",
     "PreProcessor": "NoPreProcessor",
     "param": {
-      "lr": 0.02,
-      "gamma": 0.99,
-      "hidden_layers": [64],
+      "lr": 0.005,
+      "gamma": 0.999,
+      "hidden_layers": [200, 100],
       "hidden_layers_activation": "sigmoid"
     },
     "param_range": {