| Presentation | 2020-01-10 Non-compensatory Temporal IRT with Local Variational Approximation Hiroshi Tamano, Daichi Mochihashi, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Knowledge tracing, the time varying extension of item response theory (IRT), is a task to trace learner’slatent skill states to predict whether the learner can answer a new question correctly. Due to its educational domain, knowledge tracing needs high interpretability of its prediction and explainability of its result. As for explainability, explanation with human interpretable skills is necessary, which are usually given as conjunctive conditions. Such amodel is called a non-compensatory model in multi-dimensional IRT and explanation using non-compensatory itemresponse model is desired. To realize an interpretable and explainable knowledge tracing, we propose a probabilisticmodel based on non-compensatory item response model combined with a linear dynamical system. Since it resultsin a complicated posterior on the skill states of the learners, we approximate it using a local variational distribution. We also show that our posterior adequately approximates the true posterior in artificial data, and our predictionperformance is better than two popular deep learning based knowledge tracing in ASSISTment Data. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Educational Technology / IRT / Knowledge Tracing / Kalman Filter / Variational Approximation |
| Paper # | IBISML2019-31 |
| Date of Issue | 2020-01-02 (IBISML) |
| Conference Information | |
| Committee | IBISML |
|---|---|
| Conference Date | 2020/1/9(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | ISM |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Machine learning, etc. |
| Chair | Hisashi Kashima(Kyoto Univ.) |
| Vice Chair | Masashi Sugiyama(Univ. of Tokyo) / Koji Tsuda(Univ. of Tokyo) |
| Secretary | Masashi Sugiyama(Nagoya Inst. of Tech.) / Koji Tsuda(AIST) |
| Assistant | Tomoharu Iwata(NTT) / Shigeyuki Oba(Kyoto Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Infomation-Based Induction Sciences and Machine Learning |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Non-compensatory Temporal IRT with Local Variational Approximation |
| Sub Title (in English) | |
| Keyword(1) | Educational Technology |
| Keyword(2) | IRT |
| Keyword(3) | Knowledge Tracing |
| Keyword(4) | Kalman Filter |
| Keyword(5) | Variational Approximation |
| 1st Author's Name | Hiroshi Tamano |
| 1st Author's Affiliation | NEC Data Science Laboratories/The Graduate University for Advanced Studies, SOKENDAI(NEC/SOKENDAI) |
| 2nd Author's Name | Daichi Mochihashi |
| 2nd Author's Affiliation | The Institute of Statistical Mathematics(ISM) |
| Date | 2020-01-10 |
| Paper # | IBISML2019-31 |
| Volume (vol) | vol.119 |
| Number (no) | IBISML-360 |
| Page | pp.pp.91-98(IBISML), |
| #Pages | 8 |
| Date of Issue | 2020-01-02 (IBISML) |