Optimal ensemble construction via meta-evolutionary ensembles

In this paper, we propose a meta-evolutionary approach to improve on the performance of individual classifiers. In the proposed system, individual classifiers evolve, competing to correctly classify test points, and are given extra rewards for getting difficult points right. Ensembles consisting of multiple classifiers also compete for member classifiers, and are rewarded based on their predictive performance. In this way we aim to build small-sized optimal ensembles rather than form large-sized ensembles of individually-optimized classifiers. Experimental results on 15 data sets suggest that our algorithms can generate ensembles that are more effective than single classifiers and traditional ensemble methods.     

The effects of recent practice on task switching.

Four experiments investigated the effect of recent selective practice on the cost of switching between 2 tasks afforded by letter-digit pairs: alphabet arithmetic and shape comparison. Experiments 1 and 2 found a greater cost associated with switching to the more recently practiced task: evidence that task-set inertia contributes to switching costs. Experiment 3 found this effect to be limited to trials on which a recently trained stimulus followed another such stimulus: a result problematic for all current theories of task-set priming. Experiment 4 showed that the effect of recent practice was eliminated by active preparation for a task switch: It appears that endogenous task-set preparation reduces the effects of task-set inertia. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Equivalence of models for polynomial learnability

In this paper we consider several variants of Valiant's learnability model that have appeared in the literature. We give conditions under which these models are equivalent in terms of the polynomially learnable concept classes they define. These equivalences allow comparisons of most of the existing theorems in Valiant-style learnability and show that several simplifying assumptions on polynomial learning algorithms can be made without loss of generality. We also give a useful reduction of learning problems to the problem of finding consistent hypotheses, and give comparisons and equivalences between Valiant's model and the prediction learning models of Haussler, Littlestone, and Warmuth (in “29th Annual IEEE Symposium on Foundations of Computer Science,” 1988).     

Developmental and light-dependent changes of the cytosolic chaperonin containing TCP-1 (CCT) subunits in maize seedlings, and the localization in coleoptiles

The cytosolic chaperonin containing TCP-1 (CCT) is known to keep fold cytoskeletal proteins and is involved in the proper organization of the cytoskeleton. These studies are based on the assumption that growth responses linked to structural rearrangement of the plant cytoskeleton include the action of CCT and the need for newly synthesized tubulin. The presence of the alpha- and epsilon-subunits of CCT was investigated in soluble fractions of protein extracts from maize mesocotyls and coleoptiles at distinct growth stages. The CCT-subunits, tubulins and actin decreased in the coleoptile in response to far-red light. In addition, independent from light treatment, the amount of CCT epsilon abundance declined with age in coleoptiles and mesocotyls between 2 and 4.5 days after sowing. In contrast to CCT epsilon, no significant light regulation of CCT alpha was found in the mesocotyl. In two day old, light-grown rapidly elongating coleoptiles part of the CCT alpha subunit and the bulk of actin and tubulin was found shifted into fractions of high molecular weight complexes when compared to slowly elongating, dark grown coleoptiles. In 4.5 day old, etiolated and elongating coleoptiles, part of both CCT-subunits and cytoskeleton proteins were found in fractions of high molecular weight. A complete disappearance of these polypeptides was observed in old far-red irradiated growth-arrested coleoptiles. CCT epsilon was found to be colocalized to microtubular structures and to the nucleus. We conclude from our data that abundance of CCT-subunits in soluble extracts is dependent on age and light treatment, but independent from the growth stage of mesocotyl and coleoptile.     

A new efficient dcvs circuit synthesis technique used for an improved implementation of a serial/parallel multiplier

This paper describes an efficient technique for the design of fault-secure VLSI circuits based on differential cascode voltage switch (DCVS) logic. We propose a new synthesis method for constructing DCVS circuits with a near-optimal transistor count based on binary decision diagrams (BDDs). the time and memory resources required are very low, making the technique practical even for PC-based synthesis tools. This method is the basis for a CAD tool that allows automatic synthesis of fault-secure circuits based on the DCVS technology. We finally present an improved design and implementation of a 2's complement serial/parallel multiplier as an application of the proposed technique and algorithm.     

有助于防止激光二极管毁坏的激光仿真器

激光二极管可能在数纳秒内自行毁坏，因此测试一个反馈稳定的激光二极管驱动器的响应和稳定性可能是费用很高的。图1所示仿真器电路示出了一个典型的激光二极管封装，封装内不仅有由电流1．驱动的激光二极管，而且还有一个光电二极管。激光二极管的前端面发射在外界起作用的主光束，而后端面则发射落到光电二极管上的参考光束。     

Switching between tasks of unequal familiarity: The role of stimulus-attribute and response-set selection.

It has been reported that it is harder to switch to a strong, well-practiced task from a weaker, less-practiced task than vice versa. Three experiments replicated this surprising asymmetry and investigated how it is affected by a reduction in interference between tasks. Experiment 1 progressively delayed the onset of the stimulus attribute associated with the stronger task. Experiments 2 and 3 separated the response sets of the tasks. Both manipulations reduced, without eliminating, interference of the stronger with the weaker task but reversed the asymmetry of switch costs, resulting in a larger cost of switching to the weaker task. The results are interpreted in terms of a model of the interactions between control input, task strength, and task priming. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Scalar network analysers operating at Ka- and W-bands

Scalar network analysers measuring the amplitude of the reflection/transmission coefficients of waveguide structures have been realised at Ka- and W-bands. Their accuracy derived by using standard mismatches is excellent at Ka- band and fairly good at W-band. To our knowledge this is the first instance where the accuracy of measurements obtained from a scalar network analyser was checked over full waveguide bands with the aid of standard mismatches. The reported reflectometers are precursors of vector network analysers based on six-port reflectometers. Methods aimed at increasing the signal-to-noise ratio (SNR) of the measurements have been developed, while proposals for improving the dynamic range of these network analysers are put forward.     

PAC-learnability of probabilistic deterministic finite state automata in terms of variation distance

We consider the problem of PAC-learning distributions over strings, represented by probabilistic deterministic finite automata (PDFAs). PDFAs are a probabilistic model for the generation of strings of symbols, that have been used in the context of speech and handwriting recognition, and bioinformatics. Recent work on learning PDFAs from random examples has used the KL-divergence as the error measure; here we use the variation distance. We build on recent work by Clark and Thollard, and show that the use of the variation distance allows simplifications to be made to the algorithms, and also a strengthening of the results; in particular that using the variation distance, we obtain polynomial sample size bounds that are independent of the expected length of strings.