/SAINT_plus-Knowledge-Tracing-

Implementation of [SAINT+: Integrating Temporal Features for EdNet Correctness Prediction](https://arxiv.org/abs/2010.12042)

Primary LanguagePython

SAINT_plus-Knowledge-Tracing-

SAINT+ Paper Implementaions for Riiid! Answer Correctness Prediction Competition from Kaggle.

SAINT+: Integrating Temporal Features for EdNet Correctness Prediction

This paper added an additional features to existing architecture, SAINT: Separated Self-Attentive Neural Knowledge Tracing.
SAINT: Separated Self-Attentive Neural Knowledge Tracing is from this paper Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing, its pytorch implementation is here for the same dataset given above.

Architecture of SAINT+

alt text alt text

To Run:

  - Change config.py- Datafile location, device, batch size. 
  - Saint.py file code is used while training in train.py directly. 
  - finally, run the train.py file in command line.

Citations

@misc{shin2020saint,
      title={SAINT+: Integrating Temporal Features for EdNet Correctness Prediction}, 
      author={Dongmin Shin and Yugeun Shim and Hangyeol Yu and Seewoo Lee and Byungsoo Kim and Youngduck Choi},
      year={2020},
      eprint={2010.12042},
      archivePrefix={arXiv},
      primaryClass={cs.CY}
}
@misc{choi2020appropriate,
      title={Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing}, 
      author={Youngduck Choi and Youngnam Lee and Junghyun Cho and Jineon Baek and Byungsoo Kim and Yeongmin Cha and Dongmin Shin and Chan Bae and Jaewe Heo},
      year={2020},
      eprint={2002.07033},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}