Emergence of Concepts in DNNs?
Not on the map. Why?
Abstract (arXiv)
The present paper reviews and discusses work from computer science that proposes to identify concepts in internal representations (hidden layers) of DNNs. It is examined, first, how existing methods actually identify concepts that are supposedly represented in DNNs. Second, it is discussed how conceptual spaces -- sets of concepts in internal representations -- are shaped by a tradeoff between predictive accuracy and compression. These issues are critically examined by drawing on philosophy. While there is evidence that DNNs able to represent non-trivial inferential relations between concepts, our ability to identify concepts is severely limited.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 11 of its 44 references that are in the map. First posted on arXiv 11 Nov 2022.
Builds on
11 of 44 references.
References · 11
- 2016Understanding intermediate layers using linear classifier probescited by 1,353
- 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,319
- 2014How transferable are features in deep neural networks?cited by 1,225
- 2020Concept Bottleneck Modelscited by 1,032
- 2018Reconciling modern machine learning practice and the bias-variance trade-offcited by 927
- 2017Opening the Black Box of Deep Neural Networks via Informationcited by 503
- 2020Understanding the Role of Individual Units in a Deep Neural Networkcited by 255
- 2019On Completeness-aware Concept-Based Explanations in Deep Neural Networkscited by 239
- 2020Concept Whitening for Interpretable Image Recognitioncited by 225
- 2016Information Dropout: Learning Optimal Representations Through Noisy Computationcited by 132
- 2019Achieving Conservation of Energy in Neural Network Emulators for Climate Modelingcited by 28
Built on by
Not cited yet.