fix random seeds with np.randon.RandomState()

agent/__init__.py

@@ -1,2 +1 @@
 from .util import create_agent_helper
-from .greedy import GreedyAgent

agent/greedy.py

@@ -1,58 +0,0 @@
-from recsim.agent import AbstractEpisodicRecommenderAgent
-import numpy as np
-
-class GreedyAgent(AbstractEpisodicRecommenderAgent):
-  def __init__(self, sess, observation_space, action_space, eval_mode, summary_writer):
-    super(GreedyAgent, self).__init__(action_space, summary_writer)
-    self._num_candidates = int(action_space.nvec[0])
-    self._W = np.array([[3, 1.5, 0.5]] * self._num_candidates)
-    assert self._slate_size == 1
-  def begin_episode(self, observation=None):
-    user = observation['user']
-    docs = observation['doc']
-    if 'W' in user: # use observable W
-      self._W = user['W']
-    else:
-      w = []
-      for doc_id in docs:
-        w.append(docs[doc_id])
-      self._W = np.array(w).reshape((-1, 3))
-      print("agent W:", self._W)
-    self._episode_num += 1
-    return self.step(0, observation)
-  def step(self, reward, observation):
-    docs = observation['doc']
-    user = observation['user']
-
-    base_pr = self.calc_prs(user['time'], user['last_review'], user['history'], self._W)
-    # np.exp(-last_review / np.exp(np.dot(W, x))).squeeze()
-    max_pr = -self._num_candidates
-    max_id = 0
-    for did in docs:
-      doc_id = int(did)
-      last_review = user['last_review'].copy()
-      history =  user['history'].copy()
-      last_review[doc_id] = user['time']
-      time = user['time'] + 1
-
-      history[doc_id][0] += 1
-      history[doc_id][1] += 1
-      pr1 = self.calc_prs(time, last_review, history, self._W)
-      history[doc_id][1] -= 1
-      history[doc_id][2] += 1
-      pr2 = self.calc_prs(time, last_review, history, self._W)
-      pr = (pr1 + pr2) / 2 - base_pr
-      sum_pr = np.sum(pr)
-      if sum_pr > max_pr:
-        max_pr = sum_pr
-        max_id = doc_id
-      # print("pr1", pr1)
-      # print("pr2", pr2)
-      # print("pr0", base_pr)
-    print(f"choose doc{max_id} with marginal gain {max_pr}")
-    return [max_id]
-  def calc_prs(self, train_time, last_review, history, W):
-    last_review = train_time - last_review
-    mem_param = np.exp(np.einsum('ij,ij->i', history, W))
-    pr = np.exp(-last_review / mem_param)
-    return pr

document/FlashcardDocumentSampler.py

@@ -2,13 +2,13 @@ from .FlashcardDocument import FlashcardDocument
 from recsim import document
 
 class FlashcardDocumentSampler(document.AbstractDocumentSampler):
-  def __init__(self, doc_ctor=FlashcardDocument, **kwargs):
-    super(FlashcardDocumentSampler, self).__init__(doc_ctor, **kwargs)
+  def __init__(self, doc_ctor=FlashcardDocument, seed=0, **kwargs):
+    super(FlashcardDocumentSampler, self).__init__(doc_ctor, seed, **kwargs)
     self._doc_count = 0
 
   def sample_document(self):
     doc_features = {}
     doc_features['doc_id'] = self._doc_count
-    doc_features['difficulty'] = self._rng.uniform(0, 5, (1, 3))
+    doc_features['difficulty'] = self._rng.uniform(0, 3, (1, 3))
     self._doc_count += 1
     return self._doc_ctor(**doc_features)

main.py

@@ -17,8 +17,8 @@ tf.compat.v1.disable_eager_execution()
 create_agent_fn = create_agent_helper(full_slate_q_agent.FullSlateQAgent)
 
 ltsenv = environment.Environment(
-  FlashcardUserModel(num_candidates, time_budget, slate_size),
-  FlashcardDocumentSampler(),
+  FlashcardUserModel(num_candidates, time_budget, slate_size, seed=0, sample_seed=0),
+  FlashcardDocumentSampler(seed=0),
   num_candidates,
   slate_size,
   resample_documents=False)

user/FlashcardUserModel.py

@@ -7,13 +7,14 @@ from util import eval_result
 import numpy as np
 
 class FlashcardUserModel(user.AbstractUserModel):
-  def __init__(self, num_candidates, time_budget, slate_size, seed=0):
+  def __init__(self, num_candidates, time_budget, slate_size, seed=0, sample_seed=0):
     super(FlashcardUserModel, self).__init__(
         UserResponse, UserSampler(
             UserState, num_candidates, time_budget, 
-            seed=seed
+            seed=sample_seed
         ), slate_size)
     self.choice_model = MultinomialLogitChoiceModel({})
+    self._rng = np.random.RandomState(seed)
 
   def is_terminal(self):
     terminated = self._user_state._time > self._user_state._time_budget
@@ -52,7 +53,8 @@ class FlashcardUserModel(user.AbstractUserModel):
     doc_id = doc._doc_id
     W = self._user_state._W[doc_id]
     if not W.any(): # uninitialzed
-      self._user_state._W[doc_id] = W = doc.base_difficulty * np.random.uniform(0.5, 2.0, (1, 3)) # a uniform error for each user
+      error = self._user_state._doc_error[doc_id] # a uniform error for each user
+      self._user_state._W[doc_id] = W = doc.base_difficulty * error
       print(W)
     # use exponential function to simulate whether the user recalls
     last_review = self._user_state._time - self._user_state._last_review[doc_id]
@@ -60,6 +62,6 @@ class FlashcardUserModel(user.AbstractUserModel):
 
     pr = np.exp(-last_review / np.exp(np.dot(W, x))).squeeze()
     print(f"time: {self._user_state._time}, reviewing flashcard {doc_id}, recall rate = {pr}")
-    if np.random.rand() < pr: # remembered
+    if self._rng.random_sample() < pr: # remembered
       response._recall = True
     response._pr = pr

user/UserSampler.py

@@ -7,9 +7,13 @@ class UserSampler(user.AbstractUserSampler):
                num_candidates=10,
                time_budget=60,
                **kwargs):
-    self._state_parameters = {'num_candidates': num_candidates, 'time_budget': time_budget}
     super(UserSampler, self).__init__(user_ctor, **kwargs)
-
+    doc_error = self._rng.uniform(0.5, 1.5, (num_candidates, 3))
+    self._state_parameters = {
+      'num_candidates': num_candidates, 
+      'time_budget': time_budget,
+      'doc_error': doc_error
+    }
 
   def sample_user(self):
     return self._user_ctor(**self._state_parameters)

user/UserState.py

@@ -3,13 +3,14 @@ import numpy as np
 from gym import spaces
 
 class UserState(user.AbstractUserState):
-  def __init__(self, num_candidates, time_budget):
+  def __init__(self, num_candidates, time_budget, doc_error):
     self._cards = num_candidates
     self._history = np.zeros((num_candidates, 3))
     self._last_review = np.repeat(-1.0, num_candidates)
     self._time_budget = time_budget
     self._time = 0
     self._W = np.zeros((num_candidates, 3))
+    self._doc_error = doc_error
     super(UserState, self).__init__()
   def create_observation(self):
     return {'history': self._history, 'last_review': self._last_review, 'time': self._time, 'time_budget': self._time_budget}