Sunteți pe pagina 1din 15

Theory and Application of

Cellular Automata

Nirmalya S. Maiti, Soumyabrata Ghosh, Shiladitya Munshi

CARL
Research
Group
Presentation Outline

Presentation Outline

• What is Cellular Automata (CA)?


• Background of CA Research
• What are we aiming at ?
• Our direction
• Current focus
What is CA ?

Cellular Automata: A dynamical system discrete in space and time


CA Rule: Next State Function (Ri governs next state value of ith cell)

Rule 90:
111 110 101 100 011 010 001 000
0 1 0 1 1 0 1 0

3
For 3 neighborhood CA number of possible functions is 22 = 256
n
For n cell 3 neighborhood CA number of possible Rule Vector is (256)
What is CA ?

CA Rule Vector: Global Function  State Transition Behavior


e.g.- <202 166 90 20> <6 240 60 65>
What is CA ?

Programmable Cellular Automata (PCA)


Control

Program PCA

PCA Operated by a program

Field 1 Field 2 Field 3 Field 4


Rule Vector Time steps Math/Logical Operation Branching Condition

Structure of a Program Instruction

Step No. Field 1 Field 2 Field 3 Field 4


Step 1 <184 184 184 184 184> <1> <ф > <ф >
Step 2 <232 232 232 232 232> <2> <ф > <ф >

Example of PCA Program Instruction For DCT


What is CA ?

CA Evolution Diagram

How can we utilize the computing power


of such a machine?
Background of CA Research

Background of CA Research
• John von Neumann – simulation of physical systems (1950)

• Conway’s Game of Life on two dimensional CA

• A large number researchers from various disciplines

• Stephen Wolfram started CA research in 1970


• Started Wolfram ScienceTM and developed MathematicaTM

S. Wolfram
A New Kind of Science
Wolfram Media Inc., USA, 2002, ISBN: 1579550088

• In India, Prof. P. Pal Chaudhuri started CA research in mid 80’s.


P. Pal Chaudhuri, D. Roy Chowdhury, S. Nandi, and S. Chattopadhyay,
Additive Cellular Automata Theory and Applications,
Vol. 1, IEEE Computer Society Press, USA, 1997, ISBN: 0-8186-7717-1.
Background of CA Research

Cellular Automata Research Lab (CARL)

• Started at IIT-KGP in mid 1980’s as IntelTM Research Lab (CA based


BIST for testing asynchronous circuit)

• 1998 -2003: Shifted to Bengal Engineering College, Shibpur (Now


BESU)

• 2005-2007: Sponsored by Techno India GroupTM

TM
• February 2008: Acquired by Alumnus Software Ltd
What are we aiming at ?

What are we aiming at?

• Next State function – XOR/XNOR Linear/Additive CA


.
• To simulate physical/artificial systems
– CA BIST for testing VLSI circuits
– Error Correcting Code
– Pattern Recognition

• Constraints of Linear/Additive CA

• Answer lies on use of Non-Linear CA

• No well established theory of Non-linear CA

• Analysis or synthesis tools are not available


Our Direction

Our Direction (Since 2006)

• Non-linear CA Characterization

• Characterization of programmable CA (PCA)


– Realization of given global function ( e.g.- DCT solution)

• Analysis: Identify Global function from given local function


of CA cells (CA rules)

• Synthesis: Given global function, specify the local function


of CA cells (CA rules)

• Develop CA based solution for Real Life Problems


Current Focus

Current Focus

• Non-linear CA Theory – creating new knowledge

• Encryption
– CA based AES – Software Version is ready, developing for hardware solution
based on FPGA

• Compression
– Lossless Compression of specific class of data

• Encompression (Encryption + Compression)

• Pattern Classification
– Two class classifier for a given class of data set (e.g.- classifier to detect prime
numbers)
Current Focus

CA based Advanced Encryption Standard (AES)


• AES developed by Joan Daemen and Vincent Rijmen (2002)

• AES Algorithm 128 Bit Input Text

1st 128 Bit Key


XOR
128 Bit Input Key
S-Box
0,0 0,1 0,2 0,3
Key Expansion &
Round key selection Shift Row

10th Round
1,0 1,1 1,2 1,3
2,0 2,1 2,2 2,3 Mix Column
3,0 3,1 3,2 3,3 128 X 11 Bit Key
XOR
AES Block Round Key
128 Bit Encrypted Text

• Advantages of CA based AES


• Implementation of AES in Handheld Devices
Current Focus

Biological Computing
• Lossless Compression of Specific Class of Data

– Natural Data like DNA sequences


– Absence of repeats and tandem repeats
– Available compression algorithms (like Tar, Gzip) fail

Our Approach
Transformed
Text Data CA Transformation
Data

CA Compressed
CA Rule Library
Program Data

• Identification of Homologous Protein


– CA based pattern recognition

• Immuno-informatics
– CA based Artificial Immune System
Current Focus

CA based Pattern Classification


Data from Dataset Issues:

•How to synthesize the PCA ?


PCA
A soft computing tool
After CA Transformation

New Evolved State •How to tackle real life situation with


partially specified data ?
Incorporation of probability values
DCT Module during training of the system
More 0’s More 1’s
•How to come up with neighborhood
Class 0 Class 1 size more than three?
Working with restricted domain of
extended neighborhood (Spatio-temporal
extension).
Example (Prime / Non Prime Classification)

For 7 bit data set : A PCA (k = 3) with 2 program


steps has been designed.
CARL
Research
Group

S-ar putea să vă placă și