Accelerating Relevance-Vector-Machine-Based Classification of Hyperspectral Image with Parallel Computing

Table 5

Speedup ratios of parallel RVMs as data size varies, 16 CPUs involved.

Data set	Data size	pRVM-Ensemble with different p			pRVM-Multiclass	pRVM-MatOp
		20%	40%	60%
DS-A	848 (20%)	2.96	1.55	1.13	5.7	2.3
DS-B	1697 (40%)	11.91	3.31	1.36	5.2	4.7
DS-C	2546 (60%)	17.05	4.04	1.46	5.3	6.4
DS-D	3395 (80%)	26.13	5.45	1.82	5.6	7.9
DS-E	4244 (100%)	28.10	5.52	1.87	5.4	8.8