Characterizing and Measuring the Similarity of Neural Networks with Persistent Homology
arXiv:2101.07752
Abstract
Characterizing the structural properties of neural networks is crucial yet poorly understood, and there are no well-established similarity measures between networks. In this work, we observe that neural networks can be represented as abstract simplicial complex and analyzed using their topological 'fingerprints' via Persistent Homology (PH). We then describe a PH-based representation proposed for characterizing and measuring similarity of neural networks. We empirically show the effectiveness of this representation as a descriptor of different architectures in several datasets. This approach based on Topological Data Analysis is a step towards better understanding neural networks and serves as a useful similarity measure.
References in corpus (10)
- Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
- Impact of Fully Connected Layers on Performance of Convolutional Neural Networks for Image Classification
- Statistical topological data analysis using persistence landscapes
- Similarity of Neural Network Representations Revisited
- A Topological Loss Function for Deep-Learning based Image Segmentation using Persistent Homology
- Neural Persistence: A Complexity Measure for Deep Neural Networks Using Algebraic Topology
- On Characterizing the Capacity of Neural Networks using Algebraic Topology
- Topological Machine Learning with Persistence Indicator Functions
- Path homologies of deep feedforward networks
- Capacity allocation analysis of neural networks: A tool for principled architecture design