Scribd
Upload
42 views

Genetic Programming: A Seminar On

Genetic programming is an automated method for creating computer programs to solve problems from a high-level description. It has produced 36 results comparable to human performance, including 15 that duplicate patented inventions and 2 that qualify as new patentable inventions. Genetic programming works by evolving populations of computer programs using operations inspired by natural evolution, such as reproduction, crossover, and mutation. It has been demonstrated as an "automated invention machine" capable of inventing novel solutions to problems.

Uploaded by

Zatin Gupta
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
42 views

Genetic Programming: A Seminar On

Genetic programming is an automated method for creating computer programs to solve problems from a high-level description. It has produced 36 results comparable to human performance, including 15 that duplicate patented inventions and 2 that qualify as new patentable inventions. Genetic programming works by evolving populations of computer programs using operations inspired by natural evolution, such as reproduction, crossover, and mutation. It has been demonstrated as an "automated invention machine" capable of inventing novel solutions to problems.

Uploaded by

Zatin Gupta
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 23

A Seminar

On

Genetic Programming

Presented by : PANKAJ SHARMA

ROLLNO:57

What is GENETIC PROGRAMMING ?


GP is an automated method for creating a working computer program from a high level problem statement of a problem.

GP starts from a high-level statement of what


needs to be done and automatically creates a

computer program to solve the problem.

What is the Goal?


To get computers to automatically solve the problems. It is central to : Artificial Intelligence

Machine learning
Machine Intelligence According to Turing The aim is to get machines to exhibit behaviour which if done by humans, would assume to involve the use of intelligence.

Artificial Intelligence and Machine Learning


Artificial Intelligence
It is associated with attempts to get computers to solve problems using methods relying on knowledge, logic and various analytical and mathematical methods.

Machine Learning
It is associated with attempts to get computers to solve problems that use a particular small and somewhat arbitrarily chosen set of methodologies (many of which are statistical in nature).

Machine Intelligence
It communicates the broad goal of getting computers to automatically solve problems in a human-like way. Three approaches by which human competitive machine intelligence can be achieved:1. Logic Driven search 2. Cultural search 3. Genetic or Evolutionary search

Structure of child machine = hereditary material


Changes of child machine = Mutation Natural Selection = Judgment of experimenter

What is Human Competitive?


An automatically created result is humancompetitive if it satisfies one or more of the following criteria: (A)The result was patented as an invention in the past,is an improvement over a patented invention or would qualify today as a patentable new invention.

(B)Result is equal to or better than a result accepted as new scientific result published in a journal.
(C) Result is equal to or better than a result placed into a database or archive of results maintained by a panel of scientific experts. (D)Result is publishable as a new scientific result. (E)Result is better than or equal to most recent human-created solution to a long standing problem.

(F) Result is better than or equal to a result considered as an achievement in its field when it was first discovered.

(G) The result solves a problem of indisputable difficulty in its field.


(H) Result wins a regulated competition involving human contestants (live or human-written programs) The results stand on their own merit, and are not evaluated on the basis of their mechanically produced.

Artificial to Intelligence Ratio


AI ratio of a problem-solving method is ratio of that which is delivered by the automated operation of the artificial method to the amount of intelligence that is supplied by the human applying the method to a particular problem. It is crucial to GP as it measures the value added by the artificial problem solving method.Through GP, we want to generate results with high AI ratio. e.g: Deep Blue An AI milestone. But the program had a low AI ratio.

36 Human Competitive Results Produced by GP


These results have been automatically generated by genetic programming and their results are comparable to human performance out of which: 15 GP has created an entity that either infringes or duplicates the functionality of a previously patented 20th century invention. 6 GP has done the same wrt. 21st century invention. 2 GP has created a patentable new invention.

These results have come from the fields of Molecular Biology Cellular Automata Sorting Networks Synthesis of design of topology and component sizing for the complex structures like: Analog electrical circuits Controllers

Antenna

Claimed Instance

Basis for claiming of human competitiveness


B, F

Reference

Creation of a better than classical quantum algo. For Deutsch-Jozsa early promise problem

Spector, Barnum & Bernstein, 1998

Creation of a better than B, F classical quantum algo for database search problem Creation of a soccer H playing program that won its first two games in the Robo cup 1997 comptt. Creation of a sorting network for seven items using only 16 steps. Syn. of elec. thermometer A, D

Spector, Barnum & Bernstein, Luke 1998

Genetic Programming -iii

A, G

Genetic Programming -iii

An exampleCreation of a Better-Than-Classical Quantum Algorithm for Grovers Database Search Problem

Genetic programming evolved a better-than-classical quantum algorithm for Grovers database search problem as described in a paper entitled Quantum Computing Applications of Genetic Programming.It satisfies B and F type of criteria.

High-Return of Results produced by GP


Returns are measured in terms of amount of A delivered by the method in relation to amount of I supplied by the human user. The 36 Human-Competitive results delivered a high amount of A. I is measured as amount of analysis, intelligence, information and knowledge supplied by intelligent human user prior to launching a run of genetic programming.

A Run of Genetic Program


Two types of steps are involved: (A) Preparatory steps: are problem and domain specific. They are done before the launch of problem-solving method. They establish I component of AI ratio. They include information about : Set of Terminals Set of Primitive functions Fitness Measure Parameters for controlling the run Termination criterion and method for designating result of the run.

(B) Executional steps: Are problem and domain independent. They are automatically executed during the run of problem-solving method. These are defined by flowchart of Genetic Programming. They provide A component. There is no human interference or interaction during the run of a genetic program (though human user may decide when to terminate the run).

The flowchart specifying the basic sequence of executional steps is not modified for each problem.
The flowchart shows genetic operations of : Cross Over Reproduction Mutation Architecture altering operations

Genetic Program is an Automated Invention Machine


There are 23 instances where GP has duplicated functionality of previously patented invention, infringed previously issued patent or created patentable new invention particularly in field of :

Sorting Network
Computational Circuits Electronic Thermometer etc

Patents involve :
1) Invention 2) Evolution

These involve illogical steps (which are contrary to established beliefs). Since GP does not consider these, they can unearth novel solution to problems.

16 Attributes of a System for Automatic Programming


The offspring programs should satisfy certain characteristics:
1) Starts with What needs to be done.

2) Tells How to do it.


3) Produces a computer program. 4) Automatic determination of program size. 5) Code reuse. 6) Parameterized reuse.

7) Includes internal storage.


8) Includes iterations, loops, recursions.

9) Includes wide range of programming constructs. 10) Operates in well-defined way 11) Is problem-independent 12) Has wide applicability 13) Scalability

14) Self-Organization of hierarchies


15) Automatic determination of program architecture 16) Competitive with human-produced results.

Progression of Qualitatively More Substantial Results Produced by GP


Increase in computer time results in increase in quality of the results obtained by GP. 1) Precision 2) Ability to solve problems of greater difficulty 3) Ability to solve multiple problems 4) Ability to solve problems from multiple domains

5) Results are human-competitive


6) Ability to take advantages of parallel processing

Automatic Synthesis of Parametric Topologies by means of Genetic Programming


The solution obtained by GP can be in the form of parameterized topology. It can automatically create in a single run, a general solution to a problem in the form of a graphical structure whose nodes or edges represent components and where the parameter values of the components are specified by the mathematical expressions containing free variables.

Differences between GP and AI techniques:


1) Representation: GP conducts its search for solution in a given problem space. 2) Point-to-Point transformations in search: GP transforms a set of points into another set of points 3) Roll of hill-climbing in the search: GP does not rely exclusively on Greedy algorithms. 4) Roll of determinism in the search: GP conducts search probabilistically.

5) Role of explicit knowledge base : none


6) Role of formal logic in search : none 7) Underpinnings of technique: Biologically inspired.

Application Areas for Genetic Programming


Black Art problems

Programming the Unprogrammable


Commercially Useful New Inventions

You might also like