Knowledge Transfer via Distillation of Activation Boundaries Formed by Hidden Neurons

TL;DR

This paper introduces a novel knowledge transfer method that distills the activation boundaries formed by hidden neurons in neural networks, improving the student model's ability to mimic the teacher's decision boundaries.

Contribution

It proposes a new activation boundary distillation technique with a specialized loss function, advancing knowledge transfer methods by focusing on boundary alignment.

Findings

Outperforms current state-of-the-art methods in various knowledge transfer tasks.

Effective in transferring the decision boundaries of hidden neurons.

Enhances the student model's classification performance.

Abstract

An activation boundary for a neuron refers to a separating hyperplane that determines whether the neuron is activated or deactivated. It has been long considered in neural networks that the activations of neurons, rather than their exact output values, play the most important role in forming classification friendly partitions of the hidden feature space. However, as far as we know, this aspect of neural networks has not been considered in the literature of knowledge transfer. In this paper, we propose a knowledge transfer method via distillation of activation boundaries formed by hidden neurons. For the distillation, we propose an activation transfer loss that has the minimum value when the boundaries generated by the student coincide with those by the teacher. Since the activation transfer loss is not differentiable, we design a piecewise differentiable loss approximating the activation transfer loss. By the proposed method, the student learns a separating boundary between activation region and deactivation region formed by each neuron in the teacher. Through the experiments in various aspects of knowledge transfer, it is verified that the proposed method outperforms the current state-of-the-art.