Alternating Direction Multiplier Method for Matrix -Norm Optimization in Multitask Feature Learning Problems

Table 1

Numerical results for the random problem.

b, X	ADMM-BB			IADM-MFL
(m, n, t)	ITER	TIME	RelErr	ITER	TIME	RelErr
(10000, 10, 100)	14	0.0506	2.87e − 03	32	0.1719	3.36e − 03
(20000, 10, 200)	15	0.2188	2.65e − 03	31	0.2813	2.97e − 03
(30000, 10, 300)	15	0.2344	2.98e − 03	31	0.2969	3.25e − 03
(10000, 15, 100)	18	0.0625	3.75e − 03	42	0.2500	3.49e − 03
(20000, 15, 200)	18	0.2969	3.55e − 03	41	0.3906	3.24e − 03
(30000, 15, 300)	19	0.3906	3.66e − 03	44	0.4219	3.47e − 03
(10000, 20, 100)	22	0.1406	4.78e − 03	51	0.4375	4.55e − 03
(20000, 20, 200)	22	0.3906	4.92e − 03	52	0.5613	4.27e − 03
(30000, 20, 300)	22	0.5938	5.43e − 03	54	1.3906	4.35e − 03
(10000, 25, 100)	27	0.2500	6.49e − 03	66	0.5938	4.94e − 03
(20000, 25, 200)	26	0.2969	6.40e − 03	64	0.7656	4.94e − 03
(30000, 25, 300)	26	0.7344	6.81e − 03	65	1.5156	5.11e − 03
(10000, 30, 100)	32	0.3594	9.04e − 03	80	0.7188	7.03e − 03
(20000, 30, 200)	31	0.6250	9.23e − 03	81	0.8750	6.27e − 03
(30000, 30, 300)	31	0.9688	8.47e − 03	81	1.7344	5.54e − 03