Clarify perfs of ObliqueDTPolicy compared to DTPolicy

README.md

@@ -71,7 +71,7 @@ with open("tree_halfcheetah.pkl", "rb") as f:
     clf = load(f)
 # Render
 evaluate_policy(
-    DTPolicy(clf, env),
+    ObliqueDTPolicy(clf, env),
     env=Monitor(gym.make("HalfCheetah-v4", render_mode="human")),
     render=True,
 )

docs/api.rst

@@ -23,7 +23,7 @@ Main functions
   :toctree: generated/
   :template: function.rst
 
-    interpreter.interpreter.Interpreter.train
+    interpreter.Interpreter.train
     interpreter.policies.ObliqueDTPolicy.get_oblique_data
 
 

docs/conf.py

@@ -9,7 +9,7 @@
 project = "interpreter"
 copyright = "2024, Hector Kohler"
 author = "Hector Kohler"
-release = "0.1.1"
+release = "0.1.2"
 
 # -- General configuration ---------------------------------------------------
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#general-configuration

docs/usage.md

@@ -56,7 +56,7 @@ with open("tree_halfcheetah.pkl", "rb") as f:
     clf = load(f)
 # Render
 evaluate_policy(
-    DTPolicy(clf, env),
+    ObliqueDTPolicy(clf, env),
     env=Monitor(gym.make("HalfCheetah-v4", render_mode="human")),
     render=True,
 )

interpreter/interpreter.py

@@ -79,7 +79,7 @@ def train(self, nb_iter):
             print("Fitting tree nb {} ...".format(t + 1))
             S_new, A_new = self.tree_policy.generate_data(self.env, self.data_per_iter)
             S = np.concatenate((S, S_new))
-            A = np.concatenate((A, A_new))
+            A = np.concatenate((A, self.oracle.predict(S_new)[0]))
 
             self.tree_policy.fit_tree(S, A)
             tree_reward, _ = evaluate_policy(self.tree_policy, self.env)

interpreter/policies.py

@@ -77,7 +77,7 @@ def generate_data(self, env, nb_data):
         )
         if isinstance(env.action_space, gym.spaces.Discrete):
             assert env.action_space.n == self.action_space.n
-            A = np.zeros((nb_data, 1))
+            A = np.zeros((nb_data))
         elif isinstance(env.action_space, gym.spaces.Box):
             assert env.action_space.shape == self.action_space.shape
             A = np.zeros((nb_data, self.action_space.shape[0]))

setup.py

@@ -1,6 +1,6 @@
 from setuptools import setup, find_packages
 
-__version__ = "0.1.1"
+__version__ = "0.1.2"
 
 packages = find_packages(
     exclude=[