diagNNose: A Library for Neural Activation Analysis

Jaap Jumelet


Abstract
In this paper we introduce diagNNose, an open source library for analysing the activations of deep neural networks. diagNNose contains a wide array of interpretability techniques that provide fundamental insights into the inner workings of neural networks. We demonstrate the functionality of diagNNose with a case study on subject-verb agreement within language models. diagNNose is available at https://github.com/i-machine-think/diagnnose.
Anthology ID:
2020.blackboxnlp-1.32
Volume:
Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP
Month:
November
Year:
2020
Address:
Online
Editors:
Afra Alishahi, Yonatan Belinkov, Grzegorz Chrupała, Dieuwke Hupkes, Yuval Pinter, Hassan Sajjad
Venue:
BlackboxNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
342–350
Language:
URL:
https://aclanthology.org/2020.blackboxnlp-1.32
DOI:
10.18653/v1/2020.blackboxnlp-1.32
Bibkey:
Cite (ACL):
Jaap Jumelet. 2020. diagNNose: A Library for Neural Activation Analysis. In Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP, pages 342–350, Online. Association for Computational Linguistics.
Cite (Informal):
diagNNose: A Library for Neural Activation Analysis (Jumelet, BlackboxNLP 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.blackboxnlp-1.32.pdf
Video:
 https://slideslive.com/38940638
Code
 i-machine-think/diagnnose