Learning Features with Parameter-Free Layers

TL;DR

This paper proposes replacing trainable layers with parameter-free operations like max-pool in neural networks to improve efficiency without significantly sacrificing accuracy, supported by extensive experiments on ImageNet.

Contribution

It introduces a novel approach of heavily utilizing parameter-free operations in network design, challenging the traditional reliance on trainable layers for spatial processing.

Findings

Parameter-free operations can replace trainable layers without major accuracy loss.

Networks with parameter-free layers show improved speed, fewer parameters, and reduced FLOPs.

Experimental results on ImageNet validate the efficiency benefits of the proposed design.

Abstract

Trainable layers such as convolutional building blocks are the standard network design choices by learning parameters to capture the global context through successive spatial operations. When designing an efficient network, trainable layers such as the depthwise convolution is the source of efficiency in the number of parameters and FLOPs, but there was little improvement to the model speed in practice. This paper argues that simple built-in parameter-free operations can be a favorable alternative to the efficient trainable layers replacing spatial operations in a network architecture. We aim to break the stereotype of organizing the spatial operations of building blocks into trainable layers. Extensive experimental analyses based on layer-level studies with fully-trained models and neural architecture searches are provided to investigate whether parameter-free operations such as the max-pool are functional. The studies eventually give us a simple yet effective idea for redesigning network architectures, where the parameter-free operations are heavily used as the main building block without sacrificing the model accuracy as much. Experimental results on the ImageNet dataset demonstrate that the network architectures with parameter-free operations could enjoy the advantages of further efficiency in terms of model speed, the number of the parameters, and FLOPs. Code and ImageNet pretrained models are available at https://github.com/naver-ai/PfLayer.