Choose Your Neuron: Incorporating Domain Knowledge through Neuron-Importance

By Ramprasaath R. Selvaraju, Prithvijit Chattopadhyay, Mohamed Elhoseiny, Tilak Sharma, Dhruv Batra, Devi Parikh, Stefan Lee. 2018

38 citations recorded by the source corpus.

Individual neurons in convolutional neural networks supervised for image-level classification tasks have been shown to implicitly learn semantically meaningful concepts ranging from simple textures and shapes to whole or partial objects – forming a “dictionary” of concepts acquired through the learning process. In this work we introduce a simple, efficient zero-shot learning approach based on this observation. Our approach, which we call Neuron Importance-Aware Weight Transfer (NIWT), learns to map domain knowledge about novel “unseen” classes onto this dictionary of learned concepts and then optimizes for network parameters that can effectively combine these concepts – essentially learning classifiers by discovering and composing learned semantic concepts in deep networks. Our approach shows improvements over previous approaches on the CUBirds and AWA2 generalized zero-shot learning benchmarks. We demonstrate our approach on a diverse set of semantic inputs as external domain knowledge including attributes and natural language captions. Moreover by learning inverse mappings, NIWT can provide visual and textual explanations for the predictions made by the newly learned classifiers and provide neuron names. Our code is available at https://github.com/ramprs/neuron-importance-zsl.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Devi Parikh
2. Dhruv Batra
3. Zhe Gan
4. Mohit Bansal
5. Jiasen Lu
6. Trevor Darrell
7. Stefan Lee
8. Kristen Grauman

Citations in the corpus

• Learning Without Memorizing (2018, background): “In [15], Grad-CAM maps have been shown to encode information to learn new classes, although not in an incremental setup.”
• Taking a HINT: Leveraging Explanations to Make Vision and Language Models More Grounded (2019, background): “[24] proposed an approach to learn a mapping between gradient-based importances of individual neurons within a deep network (from [25]) and class-specific domain knowledge from humans in order to learn classifiers for novel classes.”
• Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization (2016, methodology): “[48] proposed an approach that uses the gradient-based neuron importances introduced in our work, and maps it to classspecific domain knowledge from humans in order to learn classifiers for novel classes.”