Neural Echos: Depthwise Convolutional Filters Replicate Biological Receptive Fields
This work addresses improving convolutional network efficacy for vision tasks by drawing inspiration from biology, though it is incremental as it builds on existing depthwise convolution methods.
The study found that depthwise convolutional kernels in neural networks replicate biological receptive fields from the retina, and initializing models with biologically inspired weights improved accuracy on ImageNet across multiple CNN architectures.
In this study, we present evidence suggesting that depthwise convolutional kernels are effectively replicating the structural intricacies of the biological receptive fields observed in the mammalian retina. We provide analytics of trained kernels from various state-of-the-art models substantiating this evidence. Inspired by this intriguing discovery, we propose an initialization scheme that draws inspiration from the biological receptive fields. Experimental analysis of the ImageNet dataset with multiple CNN architectures featuring depthwise convolutions reveals a marked enhancement in the accuracy of the learned model when initialized with biologically derived weights. This underlies the potential for biologically inspired computational models to further our understanding of vision processing systems and to improve the efficacy of convolutional networks.