Abstract:Stochastic computing (SC) division circuits have gained importance in recent years compared to other arithmetic circuits due to their low complexity as a result of an accuracy tradeoff. Designing a division circuit is already complex in conventional binary-based hardware systems. Developing an accurate and efficient SC division circuit is an open research problem. Prior work proposed different SC division circuits by using multiplexers and JK-flip-flop units, which may require correlated or uncorrelated input bit-streams. This study is primarily centered on exploring a cost-effective and highly efficient bit-stream generator specifically designed for SC division circuits. In conjunction with this objective, we assess the performance of multiple bit-stream generators and analyze the impact of correlation on SC division. We compare different designs in terms of accuracy and hardware cost. Moreover, we discuss a low-cost and energy-efficient bit-stream generator via powers-of-2 Van der Corput (VDC) sequences. Among the tested sequence generators, our best results were achieved with VDC sequences. Our evaluation results demonstrate that the novel VDC-based design yields promising outputs, resulting in a 15.5% reduction in the area-delay product and an 18.05% saving in energy consumption for the same accuracy level compared to conventional bit-stream generators. Significantly, our investigation reveals that employing the proposed generator improves the precision compared to the state-of-the-art. We validate the proposed architecture with an image processing case study, achieving high PSNR and structural similarity values.

Design of a Stochastic Computing Architecture for the Phansalkar Algorithm

Sad-Based Stereo Vision Machine on A System-On-Programmable-Chip (Sopc)

A Parallel Bitstream Generator for Stochastic Computing

Stochastic Computing Convolution Neural Network Architecture Reinvented For Highly Efficient Artificial Intelligence Workload on Field Programmable Gate Array

From Multipliers to Integrators: A Survey of Stochastic Computing Primitives

Parallel Hybrid Stochastic-Binary-Based Neural Network Accelerators

Hybrid Stochastic-Binary Computing for Low-Latency and High-Precision Inference of CNNs

Stochastic Computing Hardware Design and Optimization for Convolutional Neutral Networks

Low-Cost and Highly-Efficient Bit-Stream Generator for Stochastic Computing Division

Stochastic Computing for Sub-0.2V Applications-Technology Evaluation and Design Optimization

Neuromorphic visual saliency implementation using stochastic computation

Optimizing Stochastic Computing for Low Latency Inference of Convolutional Neural Networks

An Energy-Efficient Mixed-Signal Parallel Multiply-Accumulate (MAC) Engine Based on Stochastic Computing

A novel high performance discrete cosine transform algorithm based on the stochastic computation

Efficient Parallel Stochastic Computing Multiply-Accumulate (MAC) Technique Using Pseudo-Sobol Bit-Streams

Efficient Non-Linear Adder for Stochastic Computing with Approximate Spatial-Temporal Sorting Network

An Accurate and Time-Efficient Subtractor by Cross Format Coding in Stochastic Computing

A Universal Approximation Method and Optimized Hardware Architectures for Arithmetic Functions Based on Stochastic Computing

A stochastic computing architecture for iterative estimation

From Stochastic to Bit Stream Computing: Accurate Implementation of Arithmetic Circuits and Applications in Neural Networks

Stochastic Synthesis for Stochastic Computing