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

In the Self-Attentive Knowledge Tracing model, past exercise interactions serve as keys and the upcoming exercise serves as the query. The authors visualized the resulting attention weights to examine which past interactions were most relevant to each prediction. They reported that exercises grouped by tag and that the model identified relationships between same-tag exercises even when those exercises were far apart in the interaction sequence.

Attention Weights Visualization (A Self-Attentive model for Knowledge Tracing)

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

San Diego State University

1. Introduction
2. Proposed method
3. Experimental setting
4. Results and Analysis
5. Conclusion and Future Work

A Self-Attentive model for Knowledge Tracing

Paper: https://arxiv.org/pdf/1907.06837.pdf

Code: https://github.com/shalini1194/SAKT/tree/master/2019-EDM

Reference for A Self-Attentive model for knowledge Tracing

Experimental Setting (A Self-Attentive model for Knowledge Tracing)

The proposed Self-Attentive model for Knowledge Tracing (SAKT) transforms student interaction sequences into a sequential modeling problem to predict whether a student will answer a question correctly. Its architecture consists of several key components: an embedding layer for interactions and exercises, a self-attention layer utilizing a scaled dot-product mechanism to assign weights to previous exercises, a feed-forward layer, residual connections, layer normalization, and a prediction layer.

Proposed Method (A Self-Attentive Model for Knowledge Tracing)

The Self-Attentive model for Knowledge Tracing (SAKT) is a transformer-based approach designed to predict student performance. Because a student's learning process heavily depends on their past experiences and acquired skills, SAKT uses a self-attention mechanism to assign weights to previously solved items, treating individual exercises as Knowledge Components (KCs). By weighting these past interactions, the model predicts the probability of a student answering a particular question correctly. In its initial proposal, SAKT demonstrated improved predictive performance over previous knowledge tracing models.

Learn Before

Related

Learn After