nicupavel / nim

nature inspired methods

NIM

Nature Inspired Methods:

• HillClimbing
• Simulated Annealing Hill Climbing
• Genetic Algorithm with variations
• Particle Swarm Optimisation

Usage

Run compare-methods.py which will generate a results file (eg results-1462734115.npy)

Run results-loader.py with the results file as parameter to generate the results table and all the graphs.

Evaluator functions

• Rastrigin
• Griewangk
• Rosenbrock
• Sixhump

http://www.geatbx.com/docu/fcnindex-01.html#P140_6155

Method Comparison

30 runs each for a maximum of 5000 evaluation functions calls

Algo	Rastrigin			Griewangk			Rosenbrock			Sixhump
	5	10	30	5	10	30	5	10	30	5	10	30
GA	2.08	7.90	162.61	0.09	0.24	4.18	3.04	12.55	111.82	-1.03	-1.02	-1.03
Std Dev	1.38	3.02	21.41	0.06	0.17	1.95	1.78	16.17	65.74	0.01	0.02	0.01
GA Hybrid	0.44	20.00	352.25	0.22	10.58	438.80	1.17	24.35	3402.52	-1.03	-1.03	-1.03
Std Dev	0.63	5.60	26.55	0.12	5.50	60.51	1.16	15.52	829.96	0.00	0.00	0.00
PSO	0.07	2.89	36.82	0.02	0.08	0.01	0.06	3.05	28.97	-1.03	-1.03	-1.03
Std Dev	0.25	1.53	9.45	0.01	0.04	0.01	0.02	1.01	14.04	0.00	0.00	0.00

Interactive Results and Graphs

https://dashboards.ly/ua-jpJYJVWT7KmDYbZkQqFRim

Non-interactive Graphs