New algorithms for budgeted learning

We explore the problem of budgeted machine learning, in which the learning algorithm has free access to the training examples’ class labels but has to pay for each attribute that is specified. This learning model is appropriate in many areas, including medical applications. We present new algorithms for choosing which attributes to purchase of which examples, based on algorithms for the multi-armed bandit problem. In addition, we also evaluate a group of algorithms based on the idea of incorporating second-order statistics into decision making. Most of our algorithms are competitive with the current state of art and performed better when the budget was highly limited (in particular, our new algorithm AbsoluteBR2). Finally, we present new heuristics for selecting an instance to purchase after the attribute is selected, instead of selecting an instance uniformly at random, which is typically done. While experimental results showed some performance improvements when using the new instance selectors, there was no consistent winner among these methods.     

Texture, strain, and phase-fraction measurements during mechanical cycling in superelastic NiTi

Superelastic NiTi was subjected to simultaneous neutron diffraction and uniaxial compressive cycling between 10 and 980 MPa. The objective was an in-situ investigation of the evolution of the stress-induced, reversible transformation between austenite and martensite, to determine the cause of the changes in the macroscopic stress-strain response with cycling. Rietveld refinement was used to analyze the neutron spectra and quantify the phase fraction, texture, and elastic strain. The average phase strain in the mechanically loaded austenite (at a given stress) remained unaltered during the 100 load-unload cycles. However, differences in both the volume fraction and texture of austenite and martensite were noted as cycling progressed, suggesting that these factors are responsible for the changes in the macroscopic stress-strain response of NiTi with mechanical cycling.     

A randomized double-blind comparison of the effects on systemic calcium homeostasis of topical calcitriol (3 micrograms/g) and calcipotriol (50 micrograms/g) in the treatment of chronic plaque psoriasis vulgaris

Calcitriol and calcipotriol are effective treatments for psoriasis, although the two have never been directly compared. We compared the efficacy and toxicity of each agent in 24 patients with moderately extensive chronic plaque psoriasis, who were randomized in double-blind fashion to apply 90 g per week of either calcitriol (3 micrograms/g) ointment or calcipotriol (50 micrograms/g) ointment over an 8-week period. Mean PASI in patients applying calcitriol fell from 13 to 8.8 (p < 0.05) and in patients applying calcipotriol from 14.9 to 4.7 (p < 0.005). The reduction was significantly greater in the calcipotriol-treated group (p < 0.05). There was a small increase in serum ionized calcium in the calcipotriol-treated group (from 1.21 mmol/L to 1.25 mmol/L, p < 0.05) but no effect on calcium homeostasis in the calcitriol group.     

Chronic gypsum fertiliser ingestion as a significant contributor to a multifactorial cattle mortality

A number of high-resolution molecular typing systems have been developed in recent years. Their availability raises the new issues of selecting the method (s) best suited for a particular purpose and interpreting and communicating typing results. Most of the currently available methods are comparative only: they allow testing of a sample of isolates for delineation of those closely related from those markedly different in genomic backgrounds. This approach is adequate for outbreak investigation, allowing determination of clonal spread in a microenvironment and identification of the source of infection. Comparative methods with sufficient resolution for most pathogens include restriction fragment-length polymorphism (RFLP), pulsed-field gel electrophoresis (PFGE), and arbitrarily primed and randomly amplified polymorphic DNA-polymerase chain reaction (PCR) analysis. For surveillance systems, monitoring clonal spread and prevalence in populations over extended periods of time requires library typing systems. These must be standardized, must have a high throughput, and must use a uniform nomenclature. Promising or validated methods include serotyping, insertion sequence fingerprinting, ribotyping, PFGE, amplified fragment-length polymorphism (AFLP), infrequent-restriction-site amplification PCR, interrepetitive element PCR typing (rep-PCR) and PCR-RFLP of polymorphic loci. PCR methods generating arrays of size-specific amplicons (AFLP, rep-PCR) can be more reproducibly analyzed by using denaturing polyacrylamide gel or capillary electrophoresis with automated laser detection. Binary probe typing systems appear optimal and should be enhanced further through use of DNA chip technology. In these systems, amplification of polymorphic regions is followed by solid-phase hybridization with a reference panel of sequence-variant specific probes. The resulting binary type results allow determination of reproducible, numeric profiles. However, interpretation and nomenclature of typing results for large-scale surveillance purposes still require a better understanding of population structure and microevolution of most microbial pathogens.     

Two Contradictory Assumptions Underlying Prognosis and Movement with Psychotherapy.

This note has relevance to the expected disagreement which may appear when various prognostic measures are compared with each other. The same remarks apply when methods of measuring movement in psychotherapy are compared with each other instead of with a criterion of improvement such as staff judgment. It is hoped that this note will stimulate examination of why different methods of judging prognosis or judging movement in psychotherapy may not correlate highly with each other, instead of permitting the conclusion that there is nothing of value in them because they do not agree with each other. We must consider whether we are basing our judgments of good prognosis on the amount of assets of the patient or on the amount of deficits of the patient. (PsycINFO Database Record (c) 2010 APA, all rights reserved)