Is Combining Classifiers with Stacking Better than Selecting the Best One?

We empirically evaluate several state-of-the-art methods for constructing ensembles of heterogeneous classifiers with stacking and show that they perform (at best) comparably to selecting the best classifier from the ensemble by cross validation. Among state-of-the-art stacking methods, stacking with probability distributions and multi-response linear regression performs best. We propose two extensions of this method, one using an extended set of meta-level features and the other using multi-response model trees to learn at the meta-level. We show that the latter extension performs better than existing stacking approaches and better than selecting the best classifier by cross validation.     

Regulation of Sertoli cell alpha 2-macroglobulin and clusterin (SGP-2) secretion by peritubular myoid cells

alpha 2-Macroglobulin and clusterin are two putative Sertoli cell secretory products; however, the regulator(s) modulating their secretion by Sertoli cells is not known. Recent studies from this laboratory have shown that the testicular alpha 2-macroglobulin, unlike its liver homologue, is not an acute-phase reactant and its concentration is not affected by acute inflammation. We sought to determine whether FSH, testosterone, and other biomolecules would affect the secretion of alpha 2-macroglobulin and clusterin by Sertoli cells as well as whether peritubular myoid cells would affect the secretion of these proteins by Sertoli cells. It was noted that Sertoli cells cultured in vitro secreted increasing amounts of alpha 2-macroglobulin and clusterin as a function of time. FSH (50-1000 ng/ml) and testosterone (10(-11)-10(-5) M) had no apparent effect on the secretion of alpha 2-macroglobulin and clusterin by Sertoli cells. Addition of interleukin-6 to Sertoli cell-enriched cultures, in doses known to stimulate alpha 2-macroglobulin secretion by hepatocytes, did not affect the alpha 2-macroglobulin secretion. However, dexamethasone at 10(-7)-10(-5) M stimulated alpha 2-macroglobulin secretion by Sertoli cells dose-dependently while the addition of interleukin-6 had no synergistic effect on dexamethasone-stimulated alpha 2-macroglobulin secretion. These findings suggest that the synthesis and/or secretion of alpha 2-macroglobulin by Sertoli cells is regulated by a mechanism distinct from that of the liver.(ABSTRACT TRUNCATED AT 250 WORDS)     

A multi-element mapping approach for size-segregated atmospheric particles using laser ablation ICP-MS combined with image analysis

This is a first attempt to measure the elemental mass loading in size-segregated aerosol particles using a laser ablation ICP-MS mapping approach in combination with image analysis software. For optimal mapping of impaction spots on foils the laser ablation ICP-MS parameters resolution, sensitivity and analysis time were critically balanced, depending on the size of the particles and the mass loading. It was shown that size-segregated particles originating from industry-influenced or urban areas could be visualized (shades of gray or “pseudocolours” representing mass loading) and digitally analyzed by comparison with a commercially available air particulate SRM (NIST 2783). Actual results for industry-influenced and urban aerosol particles show distribution profiles that are similar to these obtained with a conventional wet-chemical leaching approach (with ICP-OES or ICP-MS analysis). Also the mass loadings were in the same range although with whole-spot laser ablation ICP-MS analysis even elemental concentrations in nanoparticles could be measured whereas the leaching approach had insufficient sensitivity to measure these particles. Contrary to the use of single line or crater laser ablation ICP-MS as sometimes practiced in the literature it was found essential to map whole impaction spots due to artifacts generated by cascade impactor sampling, leading to distorted impaction spots (presence of halos or satellites).     

Variational inference for a polytomous-attribute saturated diagnostic classification model with parallel computing

Behaviormetrika - As a statistical tool to assist formative assessments in educational settings, diagnostic classification models (DCMs) have been increasingly used to provide diagnostic...