symbolic regression
Budget: ₹1,500 – ₹12,500 INR
This question will introduce you to genetic programming. Many GP systems like ECJ include symbolic regression as a tutorial example. You are to take an existing symbolic regression example (or create one if not available for your system), prepare it for the platform you are using, and execute it to get some results. Once you get it running, modify the training initialization setup (if necessary), by reading in the training points from a text file. You should have the total number of points to process and the name of the text file as parameters in the user parameter file. This will let you run your system on new data sets, without having to recompile the GP system. (Note that this modification will be useful for part B).
Run the system 10 times per experiment (each run has a new random seed), on new data you have put into a new training file. You will investigate 4 parameter variations:
1. Crossover/Mutation tests: (i) 90% crossover, 10% mutation; (ii) 100% crossover; (iii) 100% mutation.
2. Elitism test: Take the 90% c9rossover 10% mutation experiment above, and compare two variants: (iv) no elitism; (v) elitism of 2. (note that the runs in (i) are one of these variants, so there is only one new set of 10 runs here)
Run the system 10 times per experiment (each run has a new random seed), on new data you have put into a new training file. You will investigate 4 parameter variations:
1. Crossover/Mutation tests: (i) 90% crossover, 10% mutation; (ii) 100% crossover; (iii) 100% mutation.
2. Elitism test: Take the 90% c9rossover 10% mutation experiment above, and compare two variants: (iv) no elitism; (v) elitism of 2. (note that the runs in (i) are one of these variants, so there is only one new set of 10 runs here)