Transfer Learning with Binary Neural Networks

TL;DR

This paper explores transfer learning with binary neural networks, demonstrating that a single binary network trained on ImageNet can serve as an effective feature extractor for various tasks, enabling efficient hardware implementation.

Contribution

It introduces a transfer learning architecture for binary neural networks trained on ImageNet, facilitating task-specific retraining while leveraging fixed binary feature extractors.

Findings

Binary neural networks can be used as effective feature extractors for different datasets.

The proposed approach enables hardware-efficient implementation of neural networks.

A single binary network trained on ImageNet generalizes well to other tasks.

Abstract

Previous work has shown that it is possible to train deep neural networks with low precision weights and activations. In the extreme case it is even possible to constrain the network to binary values. The costly floating point multiplications are then reduced to fast logical operations. High end smart phones such as Google's Pixel 2 and Apple's iPhone X are already equipped with specialised hardware for image processing and it is very likely that other future consumer hardware will also have dedicated accelerators for deep neural networks. Binary neural networks are attractive in this case because the logical operations are very fast and efficient when implemented in hardware. We propose a transfer learning based architecture where we first train a binary network on Imagenet and then retrain part of the network for different tasks while keeping most of the network fixed. The fixed binary part could be implemented in a hardware accelerator while the last layers of the network are evaluated in software. We show that a single binary neural network trained on the Imagenet dataset can indeed be used as a feature extractor for other datasets.