TX-Ray: Quantifying and Explaining Model-Knowledge Transfer in (Un-)Supervised NLP
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Documents
- TX-Ray
Final published version, 1.43 MB, PDF document
While state-of-the-art NLP explainability (XAI) methods focus on explaining per-sample decisions in supervised end or probing tasks, this is insufficient to explain and quantify model knowledge transfer during (un-)supervised training. Thus, for TX-Ray, we modify the established computer vision explainability principle of ‘visualizing preferred inputs of neurons’ to make it usable for both NLP and for transfer analysis. This allows one to analyze, track and quantify how self- or supervised NLP models first build knowledge abstractions in pretraining (1), andthen transfer abstractions to a new domain (2), or adapt them during supervised finetuning (3) – see Fig. 1. TX-Ray expresses neurons as feature preference distributions to quantify fine-grained knowledge transfer or adaptation and guide human analysis. We find that, similar to Lottery Ticket based pruning, TX-Ray based pruning can improve test set generalization and that it can reveal how early stages of self-supervision automatically learn linguistic abstractions like parts-of-speech.
Original language | English |
---|---|
Title of host publication | Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAII) |
Editors | Jonas Peters, David Sontag |
Publisher | PMLR |
Publication date | 2020 |
Pages | 440-449 |
Publication status | Published - 2020 |
Event | 36th Conference on Uncertainty in Artificial Intelligence (UAI), - Virtuel omline Duration: 3 Aug 2020 → 6 Aug 2020 |
Conference
Conference | 36th Conference on Uncertainty in Artificial Intelligence (UAI), |
---|---|
By | Virtuel omline |
Periode | 03/08/2020 → 06/08/2020 |
Series | Proceedings of Machine Learning Research |
---|---|
Volume | 124 |
ISSN | 1938-7228 |
Number of downloads are based on statistics from Google Scholar and www.ku.dk
No data available
ID: 255044224