# Making neural networks more neural

## TL;DR

This paper introduces a biologically inspired change to neural networks that improves their ability to generalize across different types of images.

## Contribution

Hard-wiring receptive fields into the first layer of a visual DNN to enhance generalization.

## Key findings

- The modified network generalized well from photographs to sketches.
- Conventional DNNs failed to match the cross-image-type performance of the new model.

## Abstract

Deep neural networks (DNNs) are practical and effective but, despite the name, they lack biological validity. The recent study by Kang et al.1 in Patterns takes a step toward rectifying this deficit by hard-wiring receptive fields into the first layer of a visual DNN, and the authors show that their network can generalize across image types. Training on photographs, for example, resulted in good performance on sketches; conventional DNNs did not match this behavior.

Deep neural networks (DNNs) are practical and effective but, despite the name, they lack biological validity. The recent study by Kang et al. in Patterns takes a step toward rectifying this deficit by hard-wiring receptive fields into the first layer of a visual DNN, and the authors show that their network can generalize across image types. Training on photographs, for example, resulted in good performance on sketches; conventional DNNs did not match this behavior.