Margin-adaptive model selection in statistical learning

A classical condition for fast learning rates is the margin condition, first introduced by Mammen and Tsybakov. We tackle in this paper the problem of adaptivity to this condition in the context of model selection, in a general learning framework. Actually, we consider a weaker version of this condition that allows one to take into account that learning within a small model can be much easier than within a large one. Requiring this “strong margin adaptivity” makes the model selection problem more challenging. We first prove, in a general framework, that some penalization procedures (including local Rademacher complexities) exhibit this adaptivity when the models are nested. Contrary to previous results, this holds with penalties that only depend on the data. Our second main result is that strong margin adaptivity is not always possible when the models are not nested: for every model selection procedure (even a randomized one), there is a problem for which it does not demonstrate strong margin adaptivity.

Selection of project procurement methods in the consumer electronics industry : a Hong Kong study

Under pressure from both the ever increasing level of market competition and the global financial crisis, clients in consumer electronics (CE) industry are keen to understand how to choose the most appropriate procurement method and hence to improve their competitiveness. Four rounds of Delphi questionnaire survey were conducted with 12 experts in order to identify the most appropriate procurement method in the Hong Kong CE industry. Five key selection criteria in the CE industry are highlighted, including product quality, capability, price competition, flexibility and speed. This study also revealed that product quality was found to be the most important criteria for the “First type used commercially” and “Major functional improvements” projects. As for “Minor functional improvements” projects, price competition was the most crucial factor to be considered during the PP selection. These research findings provide owners with useful insights to select the procurement strategies.

The impact of organisational culture on project selection : what is the appropriate culture type?

Project selection is a complex decision making process that is not merely influenced by the technical aspects of the project. Selection of road infrastructure projects in the Indonesian public sector is generally conducted at an organisational level, which involves multiple objectives, constraints and stakeholders. Hence, a deeper understanding of the various organisational drivers that impact on such decisions, in particular organisational culture, is much needed for improving decision-making processes as it has been posited by some researchers that organisational culture can become either an enabler, or a barrier, to the process. One part of the cultural assessment undertaken as part of the research, identifies and analyses the cultural types of relevant and involved organisations in the decision making process. The organisational culture assessment instrument (OCAI) of Cameron and Quinn (2011) was utilized in this study and the data was taken from three selected provinces in Indonesia. The results can facilitate the surveyed (and similar) organisations to improve their performance by moving towards a more appropriate cultural typology that is arguably better suited to their operations and to improving their organisational processes to more closely align with their organisational vision, mission and objectives.

Quality based frame selection for video face recognition

Quality based frame selection is a crucial task in video face recognition, to both improve the recognition rate and to reduce the computational cost. In this paper we present a framework that uses a variety of cues (face symmetry, sharpness, contrast, closeness of mouth, brightness and openness of the eye) to select the highest quality facial images available in a video sequence for recognition. Normalized feature scores are fused using a neural network and frames with high quality scores are used in a Local Gabor Binary Pattern Histogram Sequence based face recognition system. Experiments on the Honda/UCSD database shows that the proposed method selects the best quality face images in the video sequence, resulting in improved recognition performance.

A robust optimization approach for index tracking problem

Index tracking is an investment approach where the primary objective is to keep portfolio return as close as possible to a target index without purchasing all index components. The main purpose is to minimize the tracking error between the returns of the selected portfolio and a benchmark. In this paper, quadratic as well as linear models are presented for minimizing the tracking error. The uncertainty is considered in the input data using a tractable robust framework that controls the level of conservatism while maintaining linearity. The linearity of the proposed robust optimization models allows a simple implementation of an ordinary optimization software package to find the optimal robust solution. The proposed model of this paper employs Morgan Stanley Capital International Index as the target index and the results are reported for six national indices including Japan, the USA, the UK, Germany, Switzerland and France. The performance of the proposed models is evaluated using several financial criteria e.g. information ratio, market ratio, Sharpe ratio and Treynor ratio. The preliminary results demonstrate that the proposed model lowers the amount of tracking error while raising values of portfolio performance measures.

FTO genotype is associated with phenotypic variability of body mass index

There is evidence across several species for genetic control of phenotypic variation of complex traits1, 2, 3, 4, such that the variance among phenotypes is genotype dependent. Understanding genetic control of variability is important in evolutionary biology, agricultural selection programmes and human medicine, yet for complex traits, no individual genetic variants associated with variance, as opposed to the mean, have been identified. Here we perform a meta-analysis of genome-wide association studies of phenotypic variation using ~170,000 samples on height and body mass index (BMI) in human populations. We report evidence that the single nucleotide polymorphism (SNP) rs7202116 at the FTO gene locus, which is known to be associated with obesity (as measured by mean BMI for each rs7202116 genotype)5, 6, 7, is also associated with phenotypic variability. We show that the results are not due to scale effects or other artefacts, and find no other experiment-wise significant evidence for effects on variability, either at loci other than FTO for BMI or at any locus for height. The difference in variance for BMI among individuals with opposite homozygous genotypes at the FTO locus is approximately 7%, corresponding to a difference of ~0.5 kilograms in the standard deviation of weight. Our results indicate that genetic variants can be discovered that are associated with variability, and that between-person variability in obesity can partly be explained by the genotype at the FTO locus. The results are consistent with reported FTO by environment interactions for BMI8, possibly mediated by DNA methylation9, 10. Our BMI results for other SNPs and our height results for all SNPs suggest that most genetic variants, including those that influence mean height or mean BMI, are not associated with phenotypic variance, or that their effects on variability are too small to detect even with samples sizes greater than 100,000.

Multi-trait assessment of early-in-life female, male and genomic measures for use in genetic selection to improve female reproductive performance of Brahman cattle

Early-in-life female and male measures with potential to be practical genetic indicators were chosen from earlier analyses and examined together with genomic measures for multi-trait use to improve female reproduction of Brahman cattle. Combinations of measures were evaluated on the genetic gains expected from selection of sires and dams for each of age at puberty (AGECL, i.e. first observation of a corpus luteum), lactation anoestrous interval in 3-year-old cows (LAI), and lifetime annual weaning rate (LAWR, i.e. the weaning rate of cows based on the number of annual matings they experienced over six possible matings). Selection was on an index of comparable records for each combination. Selection intensities were less than theoretically possible but assumed a concerted selection effort was able to be made across the Brahman breed. The results suggested that substantial genetic gains could be possible but need to be confirmed in other data. The estimated increase in LAWR in 10 years, for combinations without or with genomic measures, ranged from 8 to 12 calves weaned per 100 cows from selection of sires, and from 12 to 15 calves weaned per 100 cows from selection of sires and dams. Corresponding reductions in LAI were 60-103 days or 94-136 days, and those for AGECL were 95-125 or 141-176 days, respectively. Coat score (a measure of the sleekness or wooliness of the coat) and hip height in females, and preputial eversion and liveweight in males, were measures that may warrant wider recording for Brahman female reproduction genetic evaluation. Pregnancy-test outcomes from Matings 1 and 2 also should be recorded. Percentage normal sperm may be important to record for reducing LAI and scrotal size and serum insulin-like growth factor-I concentration in heifers at 18 months for reducing AGECL. Use of a genomic estimated breeding value (EBV) in combination with other measures added to genetic gains, especially at genomic EBV accuracies of 40%. Accuracies of genomic EBVs needed to approach 60% for the genomic EBV to be the most important contributor to gains in the combinations of measures studied.

Estimating an Aggregate Economic Genotype to Facilitate Breeding and Selection of Strawberry Genotypes in Southeast Queensland

The Queensland strawberry (Fragaria ×ananassa) breeding program in subtropical Australia aims to improve sustainable profitability for the producer. Selection must account for the relative economic importance of each trait and the genetic architecture underlying these traits in the breeding population. Our study used estimates of the influence of a trait on production costs and profitability to develop a profitability index (PI) and an economic weight (i.e., change in PI for a unit change in level of trait) for each trait. The economic weights were then combined with the breeding values for 12 plant and fruit traits on over 3000 genotypes that were represented in either the current breeding population or as progenitors in the pedigree of these individuals. The resulting linear combination (i.e., sum of economic weight × breeding value for all 12 traits) estimated the overall economic worth of each genotype as H, the aggregate economic genotype. H values were validated by comparisons among commercial cultivars and were also compared with the estimated gross margins. When the H value of ‘Festival’ was set as zero, the H values of genotypes in the pedigree ranged from –0.36 to +0.28. H was highly correlated (R2 = 0.77) with the year of selection (1945–98). The gross margins were highly linearly related (R2 > 0.98) to H values when the genotype was planted on less than 50% of available area, but the relationship was non-linear [quadratic with a maximum (R2 > 0.96)] when the planted area exceeded 50%. Additionally, with H values above zero, the variation in gross margin increased with increasing H values as the percentage of area planted to a genotype increased. High correlations among some traits allowed the omission of any one of three of the 12 traits with little or no effect on ranking (Spearman’s rank correlation 0.98 or greater). Thus, these traits may be dropped from the aggregate economic genotype, leading to either cost reductions in the breeding program or increased selection intensities for the same resources. H was efficient in identifying economically superior genotypes for breeding and deployment, but because of the non-linear relationship with gross margin, calculation of a gross margin for genotypes with high H is also necessary when cultivars are deployed across more than 50% of the available area.

Simultaneous material selection and geometry design of statically determinate trusses using continuous optimization

In this work, we explore simultaneous geometry design and material selection for statically determinate trusses by posing it as a continuous optimization problem. The underlying principles of our approach are structural optimization and Ashby’s procedure for material selection from a database. For simplicity and ease of initial implementation, only static loads are considered in this work with the intent of maximum stiffness, minimum weight/cost, and safety against failure. Safety of tensile and compression members in the truss is treated differently to prevent yield and buckling failures, respectively. Geometry variables such as lengths and orientations of members are taken to be the design variables in an assumed layout. Areas of cross-section of the members are determined to satisfy the failure constraints in each member. Along the lines of Ashby’s material indices, a new design index is derived for trusses. The design index helps in choosing the most suitable material for any geometry of the truss. Using the design index, both the design space and the material database are searched simultaneously using gradient-based optimization algorithms. The important feature of our approach is that the formulated optimization problem is continuous, although the material selection from a database is an inherently discrete problem. A few illustrative examples are included. It is observed that the method is capable of determining the optimal topology in addition to optimal geometry when the assumed layout contains more links than are necessary for optimality.

Spectral unmixing with estimated adaptive endmember index using Extended Support Vector Machine

The most difficult operation in the flood inundation mapping using optical flood images is to separate fully inundated areas from the ‘wet’ areas where trees and houses are partly covered by water. This can be referred as a typical problem the presence of mixed pixels in the images. A number of automatic information extraction image classification algorithms have been developed over the years for flood mapping using optical remote sensing images. Most classification algorithms generally, help in selecting a pixel in a particular class label with the greatest likelihood. However, these hard classification methods often fail to generate a reliable flood inundation mapping because the presence of mixed pixels in the images. To solve the mixed pixel problem advanced image processing techniques are adopted and Linear Spectral unmixing method is one of the most popular soft classification technique used for mixed pixel analysis. The good performance of linear spectral unmixing depends on two important issues, those are, the method of selecting endmembers and the method to model the endmembers for unmixing. This paper presents an improvement in the adaptive selection of endmember subset for each pixel in spectral unmixing method for reliable flood mapping. Using a fixed set of endmembers for spectral unmixing all pixels in an entire image might cause over estimation of the endmember spectra residing in a mixed pixel and hence cause reducing the performance level of spectral unmixing. Compared to this, application of estimated adaptive subset of endmembers for each pixel can decrease the residual error in unmixing results and provide a reliable output. In this current paper, it has also been proved that this proposed method can improve the accuracy of conventional linear unmixing methods and also easy to apply. Three different linear spectral unmixing methods were applied to test the improvement in unmixing results. Experiments were conducted in three different sets of Landsat-5 TM images of three different flood events in Australia to examine the method on different flooding conditions and achieved satisfactory outcomes in flood mapping.

Simultaneous Geometry Optimization and Material Selection for Truss Structures

In this work, we explore simultaneous design and material selection by posing it as an optimization problem. The underlying principles for our approach are Ashby's material selection procedure and structural optimization. For the simplicity and ease of initial implementation of the general procedure, truss structures under static load are considered in this work in view of maximum stiffness, minimum weight/cost and safety against failure. Along the lines of Ashby's material indices, a new design index is derived for trusses. This helps in choosing the most suitable material for any design of a truss. Using this, both the design space and material database are searched simultaneously using optimization algorithms. The important feature of our approach is that the formulated optimization problem is continuous even though the material selection is an inherently discrete problem.

Selection of Static VAR Compensator Location and Size for System Voltage Stability Improvement

This paper presents a methodology for selection of static VAR compensator location based on static voltage stability analysis of power systems. The analysis presented here uses the L-index of load buses, which includes voltage stability information of a normal load flow and is in the range of 0 (no load of system) to 1 (voltage collapse). An approach has been presented to select a suitable size and location of static VAR compensator in an EHV network for system voltage stability improvement. The proposed approach has been tested under simulated conditions on a few power systems and the results for a sample radial network and a 24-node equivalent EHV power network of a practical system are presented for illustration purposes. © 2000 Published by Elsevier Science S.A. All rights reserved.

Characterizing the pocketome of Mycobacterium tuberculosis and application in rationalizing polypharmacological target selection

Polypharmacology is beginning to emerge as an important concept in the field of drug discovery. However, there are no established approaches to either select appropriate target sets or design polypharmacological drugs. Here, we propose a structural-proteomics approach that utilizes the structural information of the binding sites at a genome-scale obtained through in-house algorithms to characterize the pocketome, yielding a list of ligands that can participate in various biochemical events in the mycobacterial cell. The pocket-type space is seen to be much larger than the sequence or fold-space, suggesting that variations at the site-level contribute significantly to functional repertoire of the organism. All-pair comparisons of binding sites within Mycobacterium tuberculosis (Mtb), pocket-similarity network construction and clustering result in identification of binding-site sets, each containing a group of similar binding sites, theoretically having a potential to interact with a common set of compounds. A polypharmacology index is formulated to rank targets by incorporating a measure of druggability and similarity to other pockets within the proteome. This study presents a rational approach to identify targets with polypharmacological potential along with possible drugs for repurposing, while simultaneously, obtaining clues on lead compounds for use in new drug-discovery pipelines.

Optimum Parameter Selection in Sparse Reconstruction of Frequency-Domain Optical-Coherence Tomography Signals

For a multilayered specimen, the back-scattered signal in frequency-domain optical-coherence tomography (FDOCT) is expressible as a sum of cosines, each corresponding to a change of refractive index in the specimen. Each of the cosines represent a peak in the reconstructed tomogram. We consider a truncated cosine series representation of the signal, with the constraint that the coefficients in the basis expansion be sparse. An l(2) (sum of squared errors) data error is considered with an l(1) (summation of absolute values) constraint on the coefficients. The optimization problem is solved using Weiszfeld's iteratively reweighted least squares (IRLS) algorithm. On real FDOCT data, improved results are obtained over the standard reconstruction technique with lower levels of background measurement noise and artifacts due to a strong l(1) penalty. The previous sparse tomogram reconstruction techniques in the literature proposed collecting sparse samples, necessitating a change in the data capturing process conventionally used in FDOCT. The IRLS-based method proposed in this paper does not suffer from this drawback.

Materials selection in mechanical design for microsensors and microactuators

Microsensors and microactuators are vital organs of microelectromechanical systems (MEMS), forming the interfaces between controller and environment. They are usually used for devices ranging in size at sub-millimeter or micrometer level, transforming energy between two or more domains. Presently, most of the materials used in MEMS devices belong to the silicon material system, which is the basis of the integrated circuit industry. However, new techniques are being explored and developed, and the opportunities for MEMS materials selection are getting broader. The present paper tries to apply 'performance index' to select the material best suited to a given application, in the early stage of MEMS design. The selection is based on matching performance characteristics to the requirements. A series of performance indices are given to allow a wide range comparison of materials for several typical sensing and actuating structures, and a rapid identification of candidates for a given task. (C) 2002 Elsevier Science Ltd. All rights reserved.