ConceptLens: from Pixels to Understanding

Code

• github.com/abhilekha-dalal/ConceptLens
  OfficialIn papertf★ 1

Abstract

ConceptLens is an innovative tool designed to illuminate the intricate workings of deep neural networks (DNNs) by visualizing hidden neuron activations. By integrating deep learning with symbolic methods, ConceptLens offers users a unique way to understand what triggers neuron activations and how they respond to various stimuli. The tool uses error-margin analysis to provide insights into the confidence levels of neuron activations, thereby enhancing the interpretability of DNNs. This paper presents an overview of ConceptLens, its implementation, and its application in real-time visualization of neuron activations and error margins through bar charts.

Reproductions