CactusNets: Layer Applicability as a Metric for Transfer Learning

TL;DR

This paper introduces CactusNets, a novel metric to measure how applicable learned features are to specific classes, enabling a new unsupervised learning approach inspired by the layered applicability of neural network features.

Contribution

The paper proposes a new metric for feature applicability to classes and develops CactusNet, an unsupervised learning method based on this metric.

Findings

The applicability metric effectively measures feature relevance to classes.

CactusNet demonstrates improved unsupervised learning performance.

Layer-wise applicability correlates with feature specificity.

Abstract

Deep neural networks trained over large datasets learn features that are both generic to the whole dataset, and specific to individual classes in the dataset. Learned features tend towards generic in the lower layers and specific in the higher layers of a network. Methods like fine-tuning are made possible because of the ability for one filter to apply to multiple target classes. Much like the human brain this behavior, can also be used to cluster and separate classes. However, to the best of our knowledge there is no metric for how applicable learned features are to specific classes. In this paper we propose a definition and metric for measuring the applicability of learned features to individual classes, and use this applicability metric to estimate input applicability and produce a new method of unsupervised learning we call the CactusNet.