Attentional Neural Network: Feature Selection Using Cognitive Feedback

TL;DR

The paper introduces an Attentional Neural Network framework that combines cognitive feedback with feature extraction, improving classification accuracy and digit disentanglement in noisy and complex visual tasks.

Contribution

It presents a novel, modular neural network architecture integrating top-down and bottom-up processes, enhancing robustness and versatility in visual recognition tasks.

Findings

Achieved superior or competitive accuracy on MNIST variation dataset.

Successfully disentangled overlaid digits with high success rates.

Demonstrated robustness in noisy and difficult segmentation scenarios.

Abstract

Attentional Neural Network is a new framework that integrates top-down cognitive bias and bottom-up feature extraction in one coherent architecture. The top-down influence is especially effective when dealing with high noise or difficult segmentation problems. Our system is modular and extensible. It is also easy to train and cheap to run, and yet can accommodate complex behaviors. We obtain classification accuracy better than or competitive with state of art results on the MNIST variation dataset, and successfully disentangle overlaid digits with high success rates. We view such a general purpose framework as an essential foundation for a larger system emulating the cognitive abilities of the whole brain.