maelabgrall/biocomputation_cw2

Work on genetic algorithms and particle swarm optimisation

Biologically inspired computation, coursework 2

Goal: implement a genetic algorithm (GA) and Particle Swarm Optimization (PSO), and benchmark them on a sample of functions

how to use:

(unix)

source env/bin/activate

python main.py -ga

# or

python main.py -pso

(windows)

call env\Script\activate.bat

python main.py -pso

; or

python main.py -ga

code

the code is commented a lot so it shouldn't be hard to understand it. It is mostly vectorized.

To add new benchmark function, there is an abstract class in BenchmarkFunction.py that you shoud use. You only need to implement get_name and solve(x, y). the latest should only return the result.

results:

GA:

Mean processing time in seconds (>=100 iterations)

function	10 sample	50 sample	100 sample	500 sample	1000 sample
Matyas	0.018352	0.053137	0.10877	0.834028	2.487554
Booth	0.017725	0.059882	0.110857	0.804131	2.388776
Holder Table	0.014722	0.049492	0.095943	0.769484	2.388653
Eggholder	0.032726	0.121665	0.2579	1.623453	4.061912
Himmelblau	0.016577	0.050138	0.098948	0.776121	2.395099

Mean iteration (- 100 iterations)

function	10 sample	50 sample	100 sample	500 sample	1000 sample
Matyas	28.721	10.455	3.378	0.127	0.01
Booth	27.45	16.234	11.367	0.72	0.144
Holder Table	14.209	1.602	0.755	0.09	0.005
Eggholder	122.258	135.386	160.898	107.663	63.468
Himmelblau	22.6	7.971	2.93	0.024	0

Global minimum found (% on 1000 tests)

function	10 sample	50 sample	100 sample	500 sample	1000 sample
Matyas	0.041	0.23	0.351	0.829	0.92
Booth	0.567	0.828	0.967	1	1
Holder Table	0	0	0	0	0
Eggholder	0	0	0	0	0
Himmelblau	0.829	0.997	1	1	1

PSO

Mean processing time (s)

function	10 particle	50 particle	100 particle	500 particle	1000 particle
Matyas	0.030928	0.053708	0.083029	0.311138	0.602298
Booth	0.025702	0.049388	0.080727	0.325262	0.624594
Holder Table	0.016746	0.031529	0.049911	0.200045	0.380206
Eggholder	0.017949	0.031566	0.051696	0.229794	0.439115
Himmelblau	0.020245	0.038009	0.059572	0.231943	0.443113

Mean iteration

function	10 particle	50 particle	100 particle	500 particle	1000 particle
Matyas	275.942	312.558	319.297	351.393	371
Booth	169.152	143.534	137.54	126.925	123.551
Holder Table	106.724	98.463	96.615	94.122	92.164
Eggholder	113.68	88.671	82.528	60.782	55.465
Himmelblau	133.839	127.074	122.365	117.918	116.301