EduDeepGraph
Graphs in Machine Learning (MVA) project about using Deep Learning on Graph between Educational Contents.
The aim of this project is to predict a student's answer to an educational question using a knowledge base.
With this base, we are building a similarity graph and graph convolutions for the predictions.
Dataset : lelivrescolaire dataset (questions/answers for each student)
Advisor : Julien Seznec (lelivrescolaire).
Building the graph
Jupyter notebook file
We have m(~25k) students and n(~15k) questions.
Adjacency matrix (W) / Graph :
- nodes = questions
- edges = correlations between questions (difficulty/success/spentTime mean over all students) using L1-norm
W is a n x n matrix.
Graph Convolutional Network (GCN)
Jupyter notebook file
We are using T. Kipf's paper and implementation.
Train set :
- Adjacency matrix (W, n x n matrix)
- History matrix : students' answers to the questions (H, n x m matrix, sparse matrix)