Comparing the Optimization Methods

# import libraries
import jax.numpy as jnp
from optymus.benchmark import methods_comparison

from optymus.benchmark import Himmeblau

# define the objective function
f = Himmeblau()

# define the initial point
initial_point = jnp.array([0, 5])

f.plot(n_levels=400)

comparison = methods_comparison(f, initial_point, max_iter=10)

100%|██████████| 15/15 [00:32<00:00,  2.17s/it]

comparison['results']

	method	x_opt	f_min	n_iter	time
0	Simulated Annealing	[0.32033302586059254, 2.4544921015415784]	71.7116726081363	10	0.030990
1	RMSprop	[-0.010561838806563152, -0.0031040063186497805]	170.21368514446	10	0.048882
2	Yogi	[-0.09990946851075325, 4.900401054441406]	323.167543001664	10	0.055044
3	Steepest Descent	[-2.8051180808608414, 3.131312517644538]	1.2151348386129352e-15	5	0.056394
4	Adam	[-0.10002170260609496, 4.900286739219162]	323.12696700142453	10	0.058388
5	Adagrad	[-0.050242196055189405, 4.950201257401999]	341.220381728216	10	0.058709
6	Adamax	[-0.04135715608859208, 4.958694311673667]	344.3841718367512	10	0.064358
7	BFGS	[-2.8051180871259866, 3.1313125182455286]	9.768169503529298e-19	5	0.084228
8	Univariant	[-2.8051180976907424, 3.1313125184240542]	3.74326772617301e-15	3	0.096652
9	Conjugate Gradients (fletcher_reeves)	[-2.8051180868544643, 3.131312518243001]	3.1501007730135215e-19	4	0.111740
10	CMA-ES	[2.626421342019817, 1.9941089665304592]	4.60052396806301	10	0.198422
11	Powell	[-2.805118086953216, 3.1313125182501516]	1.4604973954705433e-23	3	0.225448
12	Newton-Raphson	[2.999999999996896, 2.0000000000133182]	2.545076537982263e-21	5	0.323272
13	Particle Swarm Optimization	[2.9997535366935963, 1.9999350090290782]	2.639504986103008e-06	10	0.528195
14	Differential Evolution	[2.8904660319151163, 2.0965468814572383]	0.3828085776663928	10	1.295606

from optymus.plots import plot_optim

plot_optim(f_obj=f, x0=initial_point, comparison=comparison, n=200)