From fa494da2b14d309f557735c59199af7f3d6c8ab0 Mon Sep 17 00:00:00 2001
From: nnnyt <793313994@qq.com>
Date: Thu, 4 Mar 2021 21:17:52 +0800
Subject: [PATCH] update KLI strategy for MIRT

---
 CAT/model/IRT.py | 54 ++++++++++++++++++++++++++----------------------
 CAT/model/NCD.py | 14 ++++++++-----
 2 files changed, 38 insertions(+), 30 deletions(-)

diff --git a/CAT/model/IRT.py b/CAT/model/IRT.py
index 0693b99..273b024 100644
--- a/CAT/model/IRT.py
+++ b/CAT/model/IRT.py
@@ -1,6 +1,7 @@
 import os
 import time
 import copy
+import vegas
 import logging
 import torch
 import torch.nn as nn
@@ -192,23 +193,7 @@ def get_theta(self, student_id):
             theta of the given student
         """
         return self.model.theta.weight.data.numpy()[student_id]
-    
-    def kli(self, x, alpha, beta, pred_estimate):
-        """ The formula of KL information. Used for integral.
-        Args:
-            x: theta of student sid
-            alpha: alpha of question qid
-            beta: beta of question qid
-            pred_estimate: the estimated probability of student sid
-        Returns:
-            the formula with x
-        """
-        pred = alpha * x + beta
-        pred = 1 / (1 + np.exp(-pred))
-        q_estimate = 1  - pred_estimate
-        q = 1 - pred
-        return pred_estimate * np.log(pred_estimate / pred) + q_estimate * np.log((q_estimate / q))
-    
+
     def get_kli(self, student_id, question_id, n):
         """ get KL information
         Args:
@@ -218,19 +203,32 @@ def get_kli(self, student_id, question_id, n):
         Returns:
             v: float, KL information
         """
+        if n == 0:
+            return np.inf
         device = self.config['device']
+        dim = self.model.num_dim
         sid = torch.LongTensor([student_id]).to(device)
         qid = torch.LongTensor([question_id]).to(device)
-        theta = self.model.theta(sid).clone().detach().numpy()[0] # (10, )
-        alpha = self.model.alpha(qid).clone().detach().numpy()[0] # (10, )
+        theta = self.model.theta(sid).clone().detach().numpy()[0] # (num_dim, )
+        alpha = self.model.alpha(qid).clone().detach().numpy()[0] # (num_dim, )
         beta = self.model.beta(qid).clone().detach().numpy()[0][0] # float value
-        # pred_estimate = 1 / (1 + np.exp(-np.dot(alpha, theta.T) - beta))
         pred_estimate = self.model(sid, qid).data.numpy()[0][0] # float value
+        def kli(x):
+            """ The formula of KL information. Used for integral.
+            Args:
+                x: theta of student sid
+            """
+            pred = np.matmul(alpha.T, x) + beta
+            pred = 1 / (1 + np.exp(-pred))
+            q_estimate = 1  - pred_estimate
+            q = 1 - pred
+            return pred_estimate * np.log(pred_estimate / pred) + \
+                    q_estimate * np.log((q_estimate / q))
         c = 3
-        low = theta - c / np.sqrt(n)
-        high = theta + c / np.sqrt(n)
-        v, err = integrate.quad(self.kli, low, high, args=(alpha, beta, pred_estimate))
-        return v
+        boundaries = [[theta[i] - c / np.sqrt(n), theta[i] + c / np.sqrt(n)] for i in range(dim)]
+        integ = vegas.Integrator(boundaries)
+        result = integ(kli, nitn=10, neval=1000)
+        return result.mean
 
     def get_fisher(self, student_id, question_id):
         """ get Fisher information
@@ -250,7 +248,13 @@ def get_fisher(self, student_id, question_id):
         return fisher_info
     
     def expected_model_change(self, sid: int, qid: int, adaptest_data: AdapTestDataset):
-
+        """ get expected model change
+        Args:
+            student_id: int, student id
+            question_id: int, question id
+        Returns:
+            float, expected model change
+        """
         epochs = self.config['num_epochs']
         lr = self.config['learning_rate']
         device = self.config['device']
diff --git a/CAT/model/NCD.py b/CAT/model/NCD.py
index e223a6b..e4f6a83 100644
--- a/CAT/model/NCD.py
+++ b/CAT/model/NCD.py
@@ -93,7 +93,7 @@ def __init__(self, **config):
 
     @property
     def name(self):
-        return 'NeuralCD Model'
+        return 'Neural Cognitive Diagnosis'
 
     def init_model(self, data: Dataset):
         self.model = NCD(data.num_students, data.num_questions, data.num_concepts)
@@ -135,7 +135,7 @@ def _loss_function(self, pred, real):
     
     def adaptest_save(self, path):
         """
-        Save the model. Only save the parameters of questions(alpha, beta)
+        Save the model. Do not save the parameters for students.
         """
         model_dict = self.model.state_dict()
         model_dict = {k:v for k,v in model_dict.items() if 'student' not in k}
@@ -224,7 +224,13 @@ def evaluate(self, adaptest_data: AdapTestDataset):
         }
     
     def expected_model_change(self, sid: int, qid: int, adaptest_data: AdapTestDataset):
-
+        """ get expected model change
+        Args:
+            student_id: int, student id
+            question_id: int, question id
+        Returns:
+            float, expected model change
+        """
         epochs = self.config['num_epochs']
         lr = self.config['learning_rate']
         device = self.config['device']
@@ -252,7 +258,6 @@ def expected_model_change(self, sid: int, qid: int, adaptest_data: AdapTestDatas
             loss = self._loss_function(pred, correct)
             loss.backward()
             optimizer.step()
-            # self.model.apply_clipper()
 
         pos_weights = self.model.student_emb.weight.data.clone()
         self.model.student_emb.weight.data.copy_(original_weights)
@@ -263,7 +268,6 @@ def expected_model_change(self, sid: int, qid: int, adaptest_data: AdapTestDatas
             loss = self._loss_function(pred, wrong)
             loss.backward()
             optimizer.step()
-            # self.model.apply_clipper()
 
         neg_weights = self.model.student_emb.weight.data.clone()
         self.model.student_emb.weight.data.copy_(original_weights)