Model re-building approach

I have managed to implement model re-building. Need to run the experiments and compare the results with the initial observations.

Next step is to implement fitness inheritance approach.

Blade Experiments Re-runs

Plan for today - Tuesday 26/7/2011

Re-run article experiments on blade servers and compare the resutls.

Comparison of experiment results in the paper and the ones observed by me

Comparison

The experiments were re-run with identical settings:

• Selection - Tournament
• Crossover - Standard 2-point Crossover, rate=0.7
• Mutation - Standard Single-bit Mutation, rate=0.075
• Elitism=2

Case base	MFM-GA Experimental results (in seconds)	My observation of MFM-GA (in seconds)	MFM-GA Experimental results (SL time in seconds)	My observation of MFM-GA (SL time in seconds)
Anneal	1903	1511	122	383
Audiology	523	368	24	42
Autos	32	33	3.1	12
Bands	529	431	44.5	114
Colic	96	91	11	41
Ionosphere	133	134	12	36
Sonar	147	131	8.2	19
Vehicle	283	497	44	225

Plan for this/next week

Items to do:

1. Get understanding of code
2. Re-run experiments published in 2010 paper
3. Propose a way for model re-building

John signed up

OK, I'm signed up too.

Testing

First post

---

Let's get started