We propose iSVM - an incremental algorithm that achieves high speed in training support vector machines (SVMs) on large datasets. In the common decomposition framework, iSVM starts with a minimum working set (WS), and then iteratively selects one training example to update the WS in each optimization loop. iSVM employs a two-stage strategy in processing the training data. In the first stage, the most prominent vector among randomly sampled data is added to the WS. This stage results in an approximate SVM solution. The second stage uses temporal solutions to scan through the whole training data once again to find the remaining support vectors (SVs). We show that iSVM is especially efficient for training SVMs on applications where data size is much larger than number of SVs. On the KDD-CUP 1999 network intrusion detection dataset with nearly five millions training examples, iSVM takes less than one hour to train an SVM with 94% testing accuracy, compared to seven hours with LibSVM - one of the state-of-the-art SVM implementations. We also provide analysis and experimental comparisons between iSVM and the related algorithms.
We reported a 28-year-old man with adrenoleukodystrophy showing neurological features of olivopontocerebellar atrophy. He had a 11-year history of Addison's disease. ACTH stimulation produced no rise in the plasma cortisol level. The ratios of C24:0/C22:0, C25:0/C22:0, and C26:0/C22:0 in fatty acids of sphingomyelin from plasma were all increased. MRI showed the atrophy of brainstem and cerebellum and the abnormal hyperintense lesions of the bilateral pyramidal tracts in the brainstem and internal capsule. 99mTc-HM PAO SPECT showed hypoperfusion of the deep white matter, frontal lobes, temporal lobes, and cerebellum. We suggest that SPECT may be useful for detection of subclinical lesions in ALD.
By using a lower hybrid current drive, changes in the amplitude and phase of reflected signals at the four-waveguide grill as well as decreases in the edge plasma density measured by double probes could be monitored. The influence of the plasma density and its gradient at the grill mouth on the coupling between waveguides and a plasma was investigated using a simplified linear grill theory based on a step plus ramp density model. It was confirmed from a comparison with the theoretical analysis that changes in waveguide coupling during current drive were mainly caused by a decrease of the density in the scrape-off plasma.
