CactusNets: Layer Applicability as a Metric for Transfer Learning
This work addresses a gap in transfer learning by providing a metric for feature applicability, which could help in clustering and separating classes, though it appears incremental as it builds on existing understanding of feature behavior in networks.
The paper tackles the problem of quantifying how applicable learned features in deep neural networks are to specific classes, proposing a metric for layer applicability and introducing CactusNet as a new unsupervised learning method based on this metric.
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.