San Diego State University

1. Student Performance Prediction
2. Attention Weights Visualization
3. Ablation Study
4. Training Efficiency

Results and Analysis (A Self-Attentive model for Knowledge Tracing)

Student Performance Prediction (A Self-Attentive model for Knowledge Tracing)

In order to gain a better understanding of  which exercise from previous interactions is more relevant for the query exercise, the authors displayed attention weight for past interactions(key) and upcoming exercises.(query). Using attention weight the authors managed to have a perfect clustering of exercise tags. Additionally, it was found that their model no matter how far two exercises belonging to the same tag were , still managed to identify that they belonged to the same concept.
 

Learn Before

Related