Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Stochastic computing : techniques and applications
Gross W., Gaudet V., Springer International Publishing, New York, NY, 2019. 215 pp. Type: Book (978-3-030037-29-1)
Date Reviewed: Oct 4 2019

Stochastic computing is an approach to numerical computing that dates back to von Neumann’s work on probabilistic logics in 1952, and that requires far fewer transistors than conventional numerical processing. This economy led to extensive research on the subject in the days when computers were built from discrete components, but waned with the advent of very-large-scale integration (VLSI). However, with the growing need for highly parallel, low-power devices to support image processing and neural networks, the field is experiencing a renaissance, and this volume provides a clear, self-contained introduction.

In conventional numerical processing, a number is represented by a fixed-length base-2 string of bits. Its precision is limited by the number of bits per memory location. In stochastic computing, a number is a stream of random bits whose probability of being set to one is related to the number being processed. In the simplest case, the number is in [0,1], and the probability of a one-bit is identified with the number. The precision of the number can be extended arbitrarily by allowing the stream of bits to grow larger. Passing these streams through simple logic gates implements numerical computation. For example, to subtract a number from one, its stream is passed through a NOT gate, while the product of two bit streams is their bit-by-bit AND. The economy of stochastic computing comes from the simplicity of such gates compared with conventional numerical processors.

The first two chapters review the history of stochastic computing, while the third provides a tutorial explaining the stochastic implementation of a wide range of arithmetic operations. The third chapter reviews the tradeoff between the accuracy and correlation of different bit streams: the mapping from Boolean combinations of streams to numerical operations is valid only if the bit streams are uncorrelated with one another. The fourth chapter shows how to synthesize arbitrary polynomial functions with stochastic computing, while the fifth explores the possibility of using streams whose bits are not random in the sense of being Bernouilli sequences.

One of the major challenges in stochastic computing is generating uncorrelated streams economically. The next three chapters are devoted to various solutions to this challenge.

The last two chapters explore two application areas in greater detail: brain-inspired computing (such as deep learning and efforts to build an artificial cortex), and decoding error-correcting codes.

The individual chapters are standalone contributions by different researchers, each with its own bibliography, and there is no integrated bibliography or index. But the editors have done a good job in selecting and organizing the chapters to present a clear and convenient introduction to a computational approach that, in spite of its venerable history, will be new and inspiring to many readers.

Reviewer:  H. Van Dyke Parunak Review #: CR146716 (1912-0420)
Bookmark and Share
  Reviewer Selected
Editor Recommended
Featured Reviewer
 
 
Stochastic Analysis (D.4.8 ... )
 
 
Stochastic Programming (G.1.6 ... )
 
 
Optimization (G.1.6 )
 
 
Artificial Intelligence (I.2 )
 
 
General (G.0 )
 
 
General (I.0 )
 
  more  
Would you recommend this review?
yes
no
Other reviews under "Stochastic Analysis": Date
Selective instability: maternal effort and the evolution of gene activation and deactivation rates
Maley C., Tapscott S. Artificial Life 9(3): 317-326, 2003. Type: Article
Jun 22 2004
Financial planning via multi-stage stochastic optimization
Mulvey J., Shetty B. Computers and Operations Research 31(1): 1-20, 2004. Type: Article
Jan 15 2004
Parallel queues with resequencing
Jean-Marie A., Gün L. Journal of the ACM 40(5): 1188-1208, 1993. Type: Article
Jan 1 1995
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy