DeepShare: Sharing ReLU Across Channels and Layers for Efficient Private Inference

TL;DR

DeepShare introduces a novel method to share ReLU activations across channels and layers, significantly reducing computational costs in private inference while maintaining high accuracy in classification and segmentation tasks.

Contribution

The paper proposes a new activation module that shares DReLU operations across channels and layers, enabling more efficient private inference with state-of-the-art performance.

Findings

Reduces DReLU operations in ResNet-type networks

Achieves state-of-the-art results on classification tasks

Attains state-of-the-art performance in image segmentation

Abstract

Private Inference (PI) uses cryptographic primitives to perform privacy preserving machine learning. In this setting, the owner of the network runs inference on the data of the client without learning anything about the data and without revealing any information about the model. It has been observed that a major computational bottleneck of PI is the calculation of the gate (i.e., ReLU), so a considerable amount of effort have been devoted to reducing the number of ReLUs in a given network. We focus on the DReLU, which is the non-linear step function of the ReLU and show that one DReLU can serve many ReLU operations. We suggest a new activation module where the DReLU operation is only performed on a subset of the channels (Prototype channels), while the rest of the channels (replicate channels) replicates the DReLU of each of their neurons from the corresponding neurons in one of the prototype channels. We then extend this idea to work across different layers. We show that this formulation can drastically reduce the number of DReLU operations in resnet type network. Furthermore, our theoretical analysis shows that this new formulation can solve an extended version of the XOR problem, using just one non-linearity and two neurons, something that traditional formulations and some PI specific methods cannot achieve. We achieve new SOTA results on several classification setups, and achieve SOTA results on image segmentation.