# Introduction to stochastic computing and its challenges

@article{Hayes2015IntroductionTS, title={Introduction to stochastic computing and its challenges}, author={John P. Hayes}, journal={2015 52nd ACM/EDAC/IEEE Design Automation Conference (DAC)}, year={2015}, pages={1-3} }

We give a short overview of stochastic computing (SC) and its uses. SC computes with randomized bit-streams that loosely resemble the neural spike trains of the brain. Its key feature is the use of low-cost and low-power logic elements to implement complex numerical operations in a highly error-tolerant fashion. These advantages must be weighed against SC's inherently slow computing speed and low precision. Although studied sporadically since its invention in the 1960s, SC has regained interestâ€¦Â

## 41 Citations

Introduction to Dynamic Stochastic Computing

- Computer ScienceIEEE Circuits and Systems Magazine
- 2020

In DSC, a random bit is used to encode a single value from a digital signal and a sequence of such random bits is referred to as a dynamic stochastic sequence, well suited for implementing accumulation-based iterative algorithms such as numerical integration and gradient descent.

Introduction to Dynamic Stochastic Computing

- Computer Science
- 2020

A DSC system features a higher energy efficiency than conventional computing using a fixed-point representation with a power consumption as low as conventional SC and is potentially useful in a broad spectrum of applications including signal processing, numerical integration and machine learning.

High-speed stochastic circuits using synchronous analog pulses

- Computer Science2017 22nd Asia and South Pacific Design Automation Conference (ASP-DAC)
- 2017

This work proposes a new, high-speed, yet accurate approach for implementing Stochastic circuits that uses synchronized analog pulses as a new way of representing correlated stochastic numbers.

Time-Encoded Values for Highly Efficient Stochastic Circuits

- Computer ScienceIEEE Transactions on Very Large Scale Integration (VLSI) Systems
- 2017

Circuits synthesized with the proposed approach can work as fast and energy-efficiently as a conventional binary design while retaining the fault-tolerance and low-cost advantages of conventional stochastic designs.

COSMO: Computing with Stochastic Numbers in Memory

- Computer ScienceACM J. Emerg. Technol. Comput. Syst.
- 2022

COSMO is an architecture for computing with stochastic numbers in memory, which enables SC in memory and maximizes the performance and energy efficiency of SC by introducing several innovations.

Accurate and compact stochastic computations by exploiting correlation

- Computer ScienceTurkish J. Electr. Eng. Comput. Sci.
- 2019

Experimental results show that the methods have improved the accuracy of stochastic computation and preserved the Stochastic computing correlation without the need for conversion from SC to the conventional binary-encoded computing, and vice versa, and lower latency and lower area cost are achieved.

End-to-End Stochastic Computing

- Computer Science
- 2017

This position paper argues that embedded systems should be designed to sense, process, compute, and actuate using an approach called stochastic computing, which operates directly on the oversampled SDM representation that is a natural fit for interfacing with physical systems.

On Memory System Design for Stochastic Computing

- Computer ScienceIEEE Computer Architecture Letters
- 2018

This paper proposes a seamless stochastic system, StochMem, which features analog memory to trade the energy and area overhead of data conversion for computation accuracy, and can reduce the energy ( area) overhead by up-to 52.8% at the cost of at most 0.7% loss in computation accuracy.

An Efficient Time-based Stochastic Computing Circuitry Employing Neuron-MOS

- Computer ScienceACM Great Lakes Symposium on VLSI
- 2019

A compact and low energy circuitry of time-based stochastic computing (TBSC) have been designed, avoiding the use of complex frequency-programmable-oscillator and comparator which are exploited in the conventional TBSC circuit.

A Stochastic Computational Multi-Layer Perceptron with Backward Propagation

- Computer ScienceIEEE Transactions on Computers
- 2018

A stochastic computational multi-layer perceptron (SC-MLP) is proposed by implementing the backward propagation algorithm for updating the layer weights and the latency and energy consumption are significantly reduced.

## References

SHOWING 1-10 OF 21 REFERENCES

Survey of Stochastic Computing

- Computer ScienceTECS
- 2013

Stochastic computing is surveyed from a modern perspective where the small size, error resilience, and probabilistic features of SC may compete successfully with conventional methodologies in certain applications.

A spectral transform approach to stochastic circuits

- Computer Science2012 IEEE 30th International Conference on Computer Design (ICCD)
- 2012

A fundamental relation between stochastic circuits and spectral transforms is demonstrated and a transform approach to the analysis and synthesis of SC circuits is proposed, and it is shown that the area cost associated with Stochastic number generation can be significantly reduced.

On the Functions Realized by Stochastic Computing Circuits

- Computer ScienceACM Great Lakes Symposium on VLSI
- 2015

This paper presents, in a uniform manner and notation, what is known about the relations between the logical and stochastic behavior of stoChastic circuits and considers how correlation among input bit-streams and the presence of memory elements influences stochastics behavior.

Equivalence among stochastic logic circuits and its application

- Computer Science, Mathematics2015 52nd ACM/EDAC/IEEE Design Automation Conference (DAC)
- 2015

This work defines stochastic equivalence classes (SECs), and investigates their properties and applications, which demonstrate the computational richness of SC and leads to significant cost reductions compared to prior designs.

Stochastic Neural Computation I: Computational Elements

- Computer ScienceIEEE Trans. Computers
- 2001

The primary contribution of this paper is in introducing several state machine-based computational elements for performing sigmoid nonlinearity mappings, linear gain, and exponentiation functions, and describing an efficient method for the generation of, and conversion between, stochastic and deterministic binary signals.

Behavior of stochastic circuits under severe error conditions

- Computer Scienceit Inf. Technol.
- 2014

This paper attempts to analyze stochastic circuits under various error conditions, and to compare their behavior to that of conventional circuits under similar error conditions using probabilistic transfer matrices.

Computation on Stochastic Bit Streams Digital Image Processing Case Studies

- Computer ScienceIEEE Transactions on Very Large Scale Integration (VLSI) Systems
- 2014

This paper introduces new SCEs based on finite-state machines based on FSMs for the task of digital image processing and compares the error tolerance, hardware area, and latency of stochastic implementations to those of conventional deterministic implementations using binary radix encoding.

Stochastic circuits for real-time image-processing applications

- Computer Science2013 50th ACM/EDAC/IEEE Design Automation Conference (DAC)
- 2013

This work presents the design of several representative circuits, which demonstrate that stochastic designs can be significantly smaller, faster, more power-efficient, and more noise-tolerant than conventional ones.

An Architecture for Fault-Tolerant Computation with Stochastic Logic

- Computer ScienceIEEE Transactions on Computers
- 2011

The concept of stochastic logic is applied to a reconfigurable architecture that implements processing operations on a datapath and it is found to be much more tolerant of soft errors than conventional hardware implementations.

A Native Stochastic Computing Architecture Enabled by Memristors

- Computer ScienceIEEE Transactions on Nanotechnology
- 2014

This work simulates a memristor-based stochastic processor for gradient descent optimization, and k-means clustering, and demonstrates key advantages in energy and speed in compute-intensive, data- intensive, and probabilistic applications.