Commits

RLPy  committed 0b2e394

final swingup poleval settings

  • Participants
  • Parent commits 08bb128
  • Branches poleval

Comments (0)

Files changed (4)

File examples/swingup/poleval/ifdd.py

 
 
 def make_experiment(id=1, path="./Results/Temp/{domain}/poleval/ifdd/",
-                    discretization=16,
-                    discover_threshold=0.42878655,
+                    discretization=21,
+                    discover_threshold=1.3763480,
                     lambda_=0.0,
-                    boyan_N0=1375.098,
-                    initial_alpha=0.6329):
+                    boyan_N0=33697.309,
+                    initial_alpha=0.41957):
     logger = Logger()
     max_steps = 100000
     sparsify = 1

File examples/swingup/poleval/ifddk.py

 
 
 def make_experiment(id=1, path="./Results/Temp/{domain}/poleval/ifdd/",
-                    discretization=16,
-                    discover_threshold=0.42878655,
-                    lambda_=0.701309,
-                    kappa=1e-7,
-                    boyan_N0=1375.098,
-                    initial_alpha=0.6329):
+                    discretization=40,
+                    discover_threshold=1.377783,
+                    lambda_=0.85207829,
+                    kappa=2.647e-9,
+                    boyan_N0=36906.811,
+                    initial_alpha=0.258541):
     logger = Logger()
     max_steps = 100000
     sparsify = 1

File examples/swingup/poleval/tdlearn_rbfs.py

 
 
 def make_experiment(id=1, path="./Results/Temp/{domain}/poleval/rbfs",
-                    boyan_N0=427.1237,
-                    initial_alpha=0.05965,
+                    boyan_N0=370.8994,
+                    initial_alpha=0.0502114613,
                     resolution=9.,
-                    num_rbfs=933.,
-                    lambda_=0.435932):
+                    num_rbfs=951.0,
+                    lambda_=0.5022861):
     logger = Logger()
     max_steps = 100000
     domain = FiniteCartPoleSwingUpFriction(logger=logger)

File examples/swingup/poleval/tdlearn_tab.py

 from Experiments.PolicyEvaluationExperiment import PolicyEvaluationExperiment
 
 param_space = {'lambda_': hp.uniform("lambda_", 0., 1.),
+               'discretization': hp.quniform("discretization", 5, 40, 1),
                'boyan_N0': hp.loguniform("boyan_N0", np.log(1e1), np.log(1e5)),
                'initial_alpha': hp.loguniform("initial_alpha", np.log(5e-2), np.log(1))}
 
 
 def make_experiment(id=1, path="./Results/Temp/{domain}/poleval/tab",
-                    boyan_N0=3243.266410,
-                    initial_alpha=0.191, #0.6633,
-                    discretization=16.,
-                    lambda_=0.81140): #1953):
+                    boyan_N0=1378.86965,
+                    initial_alpha=0.0932996, #0.6633,
+                    discretization=5,
+                    lambda_=0.6433850): #1953):
     logger = Logger()
     max_steps = 100000
     domain = FiniteCartPoleSwingUpFriction(logger=logger)