FactorizeNet: Progressive Depth Factorization for Efficient Network Architecture Exploration Under Quantization Constraints

TL;DR

FactorizeNet introduces a progressive depth factorization method that allows detailed analysis of CNN layer distributions under quantization, aiding in designing efficient low-power neural networks.

Contribution

It proposes a novel progressive depth factorization strategy for CNN architecture exploration under quantization constraints, enabling layer-level insights and optimal design identification.

Findings

Enables detailed layer-wise analysis of quantized CNNs.

Facilitates efficient discovery of optimal depth-factorized architectures.

Improves understanding of efficiency-accuracy tradeoffs in low-power CNNs.

Abstract

Depth factorization and quantization have emerged as two of the principal strategies for designing efficient deep convolutional neural network (CNN) architectures tailored for low-power inference on the edge. However, there is still little detailed understanding of how different depth factorization choices affect the final, trained distributions of each layer in a CNN, particularly in the situation of quantized weights and activations. In this study, we introduce a progressive depth factorization strategy for efficient CNN architecture exploration under quantization constraints. By algorithmically increasing the granularity of depth factorization in a progressive manner, the proposed strategy enables a fine-grained, low-level analysis of layer-wise distributions. Thus enabling the gain of in-depth, layer-level insights on efficiency-accuracy tradeoffs under fixed-precision quantization. Such a progressive depth factorization strategy also enables efficient identification of the optimal depth-factorized macroarchitecture design (which we will refer to here as FactorizeNet) based on the desired efficiency-accuracy requirements.