Deep Neural Network Models Trained With A Fixed Random Classifier Transfer Better Across Domains

arXiv:2402.18614v1 Announce Type: new
Abstract: The recently discovered Neural collapse (NC) phenomenon states that the last-layer weights of Deep Neural Networks (DNN), converge to the so-called Equiangular Tight Frame (ETF) simplex, at the terminal phase of their training. This ETF geometry is equivalent to vanishing within-class variability of the last layer activations. Inspired by NC properties, we explore in this paper the transferability of DNN models trained with their last layer weight fixed according to ETF. This enforces class separation …

abstract arxiv class classifier converge cs.cv cs.lg cs.ne deep neural network dnn domains etf geometry layer network networks neural collapse neural network neural networks random terminal training transfer type