Loading...
Please wait, while we are loading the content...
Similar Documents
Solving the 8-puzzle : A Genetic Programming Approach
| Content Provider | Semantic Scholar |
|---|---|
| Author | Barnes, Jason M. Hasan, Shaddi Lee, T. Sanghwi |
| Abstract | — This paper illustrates the application of a genetic programming approach to solving the 8-puzzle, also known as the sliding block puzzle. The 8-puzzle is a 'game problem', useful for understanding concepts of machine learning in a well-defined environment. The system we have designed is broken into three functions: GP-generate, solve-8puzzle, and test-8puzzle. GP-generate uses genetic programming techniques to develop a parse tree to solve a generic 8-puzzle, while solve-8puzzle applies a specific parse tree generated by GP-generate to a specific 8-puzzle. test-8puzzle tests the solution generated by solve-8puzzle to determine if the solution is correct. We based our project on research done by Finkelstein and Markovitch on the 8-puzzle problem, specifically, our fitness function and our terminal set. Our parse trees were only able to solve the test cases to a specific point, depending on the function set used and the individual test cases. Our research determined that our learning agent was only able to partially solve the test cases. After modifications to our selection process which allowed all branches of the parse tree to be chosen for crossover with equal probabilities, we saw a dramatic decrease in the value of the heuristic. Further investigation would be necessary to reveal why the parse trees only reduced the value of the heuristic to this number rather than 0. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://sha.ddih.org/f/Barnes-Hasan-Lee-Project-3.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |