A Self-Adjusting Spectral Conjugate Gradient Method for Large-Scale Unconstrained Optimization

Table 1

Numerical results for PRP⁺ method.

Function	𝑛	NI	Nfg	𝑇 (0.01 S)	‖𝑔(𝑥)‖∞
Quadratic QF2	10000	2227	2885	2016	9.98𝐸−07
Extended EP1	10000	4	7	3	6.09𝐸−13
Extended Tridiagonal 2	10000	39	98	47	9.29𝐸−07
ARGLINA	10000	5	15	4	1.95𝐸−07
ARWHEAD	10000	7	14	21	3.10𝐸−07
BDQRTIC	5000	157	720	526	1.47𝐸−04
BDEXP	5000	6	8	7	1.72𝐸−07
BRYBND	5000	5	11	1215	2.60𝐸−07
COSINE	10000	21	45	39	9.60𝐸−07
CRAGGLVY	10000	129	250	444	5.35𝐸−06
DIXMAANA	10000	6	12	19	4.20𝐸−07
DIXMAANB	10000	8	16	26	6.72𝐸−07
DIXMAANC	10000	11	23	38	3.38𝐸−08
DIXMAAND	10000	13	29	44	1.32𝐸−07
DIXMAANE	5000	558	799	712	9.96𝐸−07
DIXMAANF	5000	558	598	525	8.65𝐸−07
DIXMAANG	5000	519	784	684	5.99𝐸−07
DIXMAANH	5000	379	3488	2469	8.97𝐸−07
DIXMAANI	5000	593	854	755	7.51𝐸−07
DIXMAANJ	5000	492	751	651	5.08𝐸−07
DIXMAANK	5000	653	979	863	9.61𝐸−07
DQDRTIC	10000	11	23	19	9.59𝐸−08
DQRTIC	10000	33	57	42	3.44𝐸−07
EDENSCH	10000	26	90	78	9.10𝐸−06
EG2	10000	209	1426	473	1.10𝐸−03
ENGVAL1	10000	30	93	21	1.37𝐸−06
EXTROSNB	10000	29	63	25	3.77𝐸−08
FREUROTH	10000	61	145	81	2.33𝐸−07
LIARWHD	10000	25	49	36	6.54𝐸−09
NONDIA	10000	9	17	17	1.03𝐸−09
NONDQUAR	5000	1786	3258	1752	7.96𝐸−07
NONSCOMP	10000	5001	6799	9751	3.47𝐸−06