AICTE Sponsored

International Conference on
Advances in
Computer
Engineering & Communication
Technology
(ICACET-2021)
Morressier Paper Id: 176
Paper Title: Stochastic Computing Solutions Challenges and Application
Author: Surya Prakash Sangisetti

22nd, 23rd October 2021

 Contents
 Abstract
 Introduction
 Methodology for Stochastic Computing
 Implementation Technique
 Conclusion
 References

ICACET 2021 2
 Abstract

 Modern-day applications need high speed and power with low power consumption and
a small integration area. So this system will be consumed less power.
 While loaded computing contributes high hardware cost, high speed. Stochastic
computing accuracy is less than binary computing circuits. They are having some
implementation basic operations of stochastic computing. They are complementary
operation, multiplication operation, addition operation, subtraction operation, and
division operation.
 These operations having some logic gates i.e.., AND gate, NOT gate, XOR gate, an
XNOR gate. And implement the hardware Tripartite synapse.

ICACET 2021 3
 Introduction

Recently IT Technology such as image processing classifiers becoming more complex and requiring
more energy. Which is probabilistic. Stochastic computing is one of the approximate computing .

•  It is useful for energy efficiency.  Stochastic computing is independently developed by two
groups 

•  Stochastic computing uses a sequence of bits which is called bitstream. The advantages of
stochastic computing are small size, low power, accuracy, high error tolerance. 

•  The main advantage of stochastic computing is it enables the implementation of complex

arithmetic functions. 

• The drawbacks of stochastic computing are low speed, low accuracy, and low precision 

ICACET 2021 4
 • The Stochastic Number generators are large due that it occupies the maximum area in
stochastic computing.
 
• In unipolar and bipolar format, the operation of addition is performed by a multiplexer .

•  Due to the nature of stochastic computing, the sequence of numbers cannot be greater
than one. There are two alternatives to operate the delayed version of the input signals.

•  They are converting the input into a stochastic bitstream and the input signal first passes
through the delay line 

ICACET 2021 5
 Methodology For Stochastic Computing

Conversion Blocks: To use stochastic logic, it is required to convert a binary into its   
corresponding stochastic number and stochastic into its corresponding binary number. It      consists
of the comparator and random number generator. The probability distribution function and density
function for this block is given                                                 
fXx={12n       for 0≤X< 2n                        0         for X≥2n                                                 1

FXx=Px<b= {b2n           for 0≤b <2n 1       for b≤2n                                 (2) 

                        
Pulse to binary converter consists of a simple counter. The output of this block is given as 
  PNX= N X .PX.(1-P)N_X                                                                          (3)

ICACET 2021 6
 Figure 1: Block diagram for pulse binary converter

ICACET 2021 7
  A binary counter is a digital circuit consists of clock input and several count outputs
which gives the number of clock cycles. The output may vary either on falling or rising
clock edges. A comparator is a device for comparing an object with a similar object or
with a standard measure

  Figure2: Arithmetic operations in stochastic implementation

Complementary
Multiplication Addition
Operation
Operation Operation

Division Subtraction
Operation Operation

ICACET 2021 8
  Figure 3. Optimized Tripartie Synapse

LIF
Astrocyte neuron

Synapse

Input Output
Spiking spiking

ICACET 2021 9
 Implementation Technique
 Implementation Technique
Stochastic computing is a technique to implement complex functions at a low cost.
Stochastic computing works in different number formats. The number format travels
through stochastic designs with high accuracy, less power consumption, and less
complexity.
 It consists of Complementary operation, Multiplication operation, Addition operation,
Division operation, Subtraction operation. Complementary operation is using simple
NOT gate to perform the unipolar and bipolar format.

a
a b c
b

NOT GATE AND GATE

10
  The number format travels through stochastic designs with high accuracy, less power
consumption, and less complexity. the Block diagram for the pulse binary converter.
 It consists of a Stochastic Bitstream, System clock, Binary counter, Evaluation Period,
Comparator, and Binary output.
 It consists of Complementary operation, Multiplication operation, Addition operation,
Division operation, Subtraction operation. Complementary operation is using simple
NOT gate to perform the unipolar and bipolar format.

ICACET 2021 11
 Conclusion


Stochastic computing is a technique to implement complex functions at a low cost.
Stochastic computing works in different number formats.
 The number format travels through stochastic designs with high accuracy, less power
consumption, and less complexity.
 It consists of a Stochastic Bitstream, System clock, Binary counter, Evaluation Period,
Comparator, and Binary output. shows the Stochastic implementation of basic
arithmetic operations.
 It consists of Complementary operation, Multiplication operation, Addition operation,
Division operation, Subtraction operation. Complementary operation is using simple
NOT gate to perform the unipolar and bipolar format.

ICACET 2021 12
 Reference
 1. Alaghi, A., Qian, W., &amp; Hayes, J. P. (2017). The promise and challenge of
stochastic computing. IEEE Transactions on Computer-Aided Design of
Integrated Circuits and Systems, 37(8), 1515-1531.
 2. Akhtar, R., &amp; Khanday, F. A. (2018, October). Stochastic Computing: Systems,

Applications, Challenges and Solutions. In 2018 3rd International Conference on

Communication and Electronics Systems (ICCES) (pp. 722-727). IEEE.
 3. Lee, V. T., Alaghi, A., &amp; Ceze, L. (2018, March). Correlation manipulating

circuits for stochastic computing. In 2018 Design, Automation &amp; Test in Europe
Conference &amp; Exhibition (DATE) (pp. 1417-1422). IEEE.
 4. Veerendra, G., Swaroop, R., Dattu, D. S., Jyothi, C. A., &amp; Singh, M. K. (2021).

Detecting plant Diseases, quantifying and classifying digital image processing

techniques. Materials Today: Proceedings.

ICACET 2021 13
  5. Li, S., Glova, A. O., Hu, X., Gu, P., Niu, D., Malladi, K. T., ... &amp; Xie, Y. (2018,
October). Scope: A stochastic computing engine for dram-based in-situ accelerator.
In 2018 51st Annual IEEE/ACM International Symposium on Microarchitecture
(MICRO) (pp. 696-709). IEEE.

 6. Balaji, V. N., Srinivas, P. B., &amp; Singh, M. K. (2021). Neuromorphic

advancements architecture design and its implementations technique. Materials
Today: Proceedings.

ICACET 2021 14
 Thanku You
- Surya

ICACET 2021 15