GEO: Generation and Execution Optimized Stochastic Computing Accelerator for Neural Networks

Stochastic computing (SC) has seen a renaissance in recent years as a means for machine learning acceleration due to its compact arithmetic and approximation properties. Still, SC accuracy remains an issue, with prior works either not fully utilizing the computational density or suffering from significant accuracy losses. In this work, we propose GEO - Generation and Execution Optimized Stochastic Computing Accelerator for Neural Networks, which optimizes stream generation and execution components of SC, and bridges the accuracy gap between stochastic computing and fixed-point neural networks. It improves accuracy by coupling controlled stream sharing with training and balancing OR and binary accumulations. GEO further optimizes the SC execution through progressive shadow buffering and architectural optimizations. GEO can improve accuracy compared to state-of-the-art SC by 2.2-4.0% points while being up to 4.4X faster and 5.3X more energy efficient. GEO eliminates the accuracy gap between SC and fixed-point architectures while delivering up to 5.6X higher throughput and 2.6X lower energy.

I. Introduction

The rapid growth of deep learning in the past decade has created an immense demand for computing power at both the cloud and edge. Multiple algorithmic, architectural, and circuit approaches have been proposed to meet this demand. Among those, stochastic computing (SC) has been enjoying a renaissance in deep learning acceleration for latency-, energy-, and cost-constrained devices [1]–[5]. It offers a very compact computing footprint, enabling high levels of parallelism and data reuse not achievable using conventional floating- or fixed-point architectures [5]. Its approximate nature synergizes well with neural networks' inherent error-tolerant properties, enabling new axes of accuracy and performance tradeoffs [3], [5], [6].