Saccade Attention Networks: Using Transfer Learning of Attention to Reduce Network Sizes

TL;DR

This paper introduces Saccade Attention Networks that leverage transfer learning of attention to identify key features, significantly reducing network size and computation while maintaining performance.

Contribution

It proposes a novel method to pre-process images using learned saccade attention, enabling substantial reduction in network size and computational cost.

Findings

Reduced calculations by nearly 80%

Achieved similar results with smaller networks

Utilized transfer learning to identify key features

Abstract

One of the limitations of transformer networks is the sequence length due to the quadratic nature of the attention matrix. Classical self attention uses the entire sequence length, however, the actual attention being used is sparse. Humans use a form of sparse attention when analyzing an image or scene called saccades. Focusing on key features greatly reduces computation time. By using a network (Saccade Attention Network) to learn where to attend from a large pre-trained model, we can use it to pre-process images and greatly reduce network size by reducing the input sequence length to just the key features being attended to. Our results indicate that you can reduce calculations by close to 80% and produce similar results.