I have learnt the basics of genetic algorithms and how they are implemented in general. However, I would like to enrich my knowledge of this field. For example, to develop representation, selection and crossover methods not just apply existing patterns. What books, courses or papers do you suggest for me to use for learning?
Try this: An Introduction to Genetic Algorithms, by Melanie Mitchell
It provides a great introduction to the subject, and contains everything you need to know to start writing GA programs..
Also u may try this one, "multi-objective Optimization using Evolutionary Algorithms" by Kalyanmoy Deb :)
In addition to the good books suggested above, I recommend this book
Haupt, R. L., & Haupt, S. E. (2004). Practical Genetic Algorithms: Wiley.
Another good book
Sivanandam, S. N., & Deepa, S. N. (2010). Introduction to Genetic Algorithms: Springer.
For fundamentals follow Goldberg. For mathematical background see the original text by John Holland and Koza. There is also an article in Scientific American [1990] by Holland. The convergence of the algorithm is guided by Schemata Theorem , Rest is only various implementations.
Particularly the SA article John Holland is extremely informative. Moreover there is a review article in Science , I have forgotten the year , may be 1991 or 92. To get a comparative view one must also learn Simulated Annealing Algorithm and must understand the Annealing Theorem.
A lot of the early books (Goldberg, Mitchell etc.) are very readable. However, bear in mind that this is a very active research field, so I would urge you to consider more recent material. For instance, it is now recognised that binary encodings and roulette wheel selection (which are standard in earlier books) are often sub-optimal and have been largely replaced with real-valued encodings, Gaussian mutation, tournament selection etc. Likewise, ideas like the schema theorem and implicit parallelism (central to Goldberg's book) are rarely discussed these days. I would suggest Sean Luke's Essentials of Metaheuristics for an up-to-date discussion of GAs and related methods: see http://cs.gmu.edu/~sean/book/metaheuristics/ (note, this is free to download).
@ Michael
Good points. I tried the link but it did not work. Could you check and post it again. thank you.
I realise I didn't really answer the original question - which is about developing new representations and operators for GAs. This is something that a lot of people have tried to do over the years, with varying degrees of success. The issue is that a representation or operator that is good for one problem is not necessarily good for others - as suggested by the No Free Lunch theorem. However, some interesting ideas you might look at include homologous crossover, path relinking, directed mutation, linkage learning and deterministic crowding. Google Scholar is a good place to start.
I would like to suggest the review article by Stephanie Forrest in Science for both beginners and researchers.
The title is: Genetic algorithms: principles of natural selection applied to computation, Science, 13 August 1993, Vol. 261 no. 5123 pp. 872-878
Yes Rahul ,you are correct. In fact , after one of my students submitted his masters dissertation on Genetic Programming in 1998 , i had no more contact with the latest papers. I switched over to the emerging field of High Performance Computing. My latest research interest is in Compressed Sensing , the emerging sampling paradigm which is slated to change the digital world surrounding us.
Anyway , still I remember the enthusiasm generated by GAs in 1909s.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages