Do pseudo-absence selection strategies influence species distribution models and their predictions? An information-theoretic approach based on simulated data

Background Multiple logistic regression is precluded from many practical applications in ecology that aim to predict the geographic distributions of species because it requires absence data, which are rarely available or are unreliable. In order to use multiple logistic regression, many studies have simulated "pseudo-absences" through a number of strategies, but it is unknown how the choice of strategy influences models and their geographic predictions of species. In this paper we evaluate the effect of several prevailing pseudo-absence strategies on the predictions of the geographic distribution of a virtual species whose "true" distribution and relationship to three environmental predictors was predefined. We evaluated the effect of using a) real absences b) pseudo-absences selected randomly from the background and c) two-step approaches: pseudo-absences selected from low suitability areas predicted by either Ecological Niche Factor Analysis: (ENFA) or BIOCLIM. We compared how the choice of pseudo-absence strategy affected model fit, predictive power, and information-theoretic model selection results. Results Models built with true absences had the best predictive power, best discriminatory power, and the "true" model (the one that contained the correct predictors) was supported by the data according to AIC, as expected. Models based on random pseudo-absences had among the lowest fit, but yielded the second highest AUC value (0.97), and the "true" model was also supported by the data. Models based on two-step approaches had intermediate fit, the lowest predictive power, and the "true" model was not supported by the data. Conclusion If ecologists wish to build parsimonious GLM models that will allow them to make robust predictions, a reasonable approach is to use a large number of randomly selected pseudo-absences, and perform model selection based on an information theoretic approach. However, the resulting models can be expected to have limited fit.

Selection strategies of segregant soybean populations for resistance to Asian rust

The objective of this work was to identify the best selection strategies for the more promising parental combinations to obtain lines with good resistance to soybean Asian rust (Phakopsora pachyrhizi). Two experiments were carried out in the field during the 2006/2007 and 2007/2008 growing seasons, to determine the percentage of infected leaf area of individual plants of five parents and their segregant F2 and F3 populations. The data obtained indicates that additive genetic variance predominates in the control of soybean resistance to Asian rust, and that the year and time of assessment do not significantly influence the estimates of the genetic parameters obtained. The narrow-sense heritability (h²r) ranged from 23.12 to 55.83%, and indicates the possibility of successful selection of resistant individuals in the early generations of the breeding program. All the procedures used to select the most promising populations to generate superior inbred lines for resistance to P. pachyrhizi presented similar results and identified the BR01-18437 x BRS 232 population as the best for inbred line selection.

On Arbitrary Selection Strategies for Basic Superposition

For first-order Horn clauses without equality, resolution is complete with an arbitrary selection of a single literal in each clause [dN 96]. Here we extend this result to the case of clauses with equality for superposition-based inference systems. Our result is a generalization of the result given in [BG 01]. We answer their question about the completeness of a superposition-based system for general clauses with an arbitrary selection strategy, provided there exists a refutation without applications of the factoring inference rule.

Selection strategies for dairy buffaloes: economic and genetic consequences

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Breeding strategies for recurrent selection of maize

The objectives of this work were to analyze theoretical genetic gains of maize due to recurrent selection among full-sib and half-sib families, obtained by Design I, Full-Sib Design and Half-Sib Design, and genotypic variability and gene loss with long term selection. The designs were evaluated by simulation, based on average estimated gains after ten selection cycles. The simulation process was based on seven gene systems with ten genes (with distinct degrees of dominance), three population classes (with different gene frequencies), under three environmental conditions (heritability values), and four selection strategies. Each combination was repeated ten times, amounting to 25, 200 simulations. Full-sib selection is generally more efficient than half-sib selection, mainly with favorable dominant genes. The use of full-sib families derived by Design I is generally more efficient than using progenies obtained by Full-Sib Design. Using Design I with 50 males and 200 females (effective size of 160) did not result in improved populations with minimum genotypic variability. In the populations with lower effective size (160 and 400) the loss of favorable genes was restricted to recessive genes with reduced frequencies.

Strategies in identifying individuals in a segregant population of common bean and implications of genotype x environment interaction in the success of selection

The objective of this study was to compare the BLUP selection method with different selection strategies in F-2:4 and assess the efficiency of this method on the early choice of the best common bean (Phaseolus vulgaris) lines. Fifty-one F-2:4 progenies were produced from a cross between the CVIII8511 x RP-26 lines. A randomized block design was used with 20 replications and one-plant field plots. Character data on plant architecture and grain yield were obtained and then the sum of the standardized variables was estimated for simultaneous selection of both traits. Analysis was carried out by mixed models (BLUP) and the least squares method to compare different selection strategies, like mass selection, stratified mass selection and between and within progeny selection. The progenies selected by BLUP were assessed in advanced generations, always selecting the greatest and smallest sum of the standardized variables. Analyses by the least squares method and BLUP procedure ranked the progenies in the same way. The coincidence of the individuals identified by BLUP and between and within progeny selection was high and of the greatest magnitude when BLUP was compared with mass selection. Although BLUP is the best estimator of genotypic value, its efficiency in the response to long term selection is not different from any of the other methods, because it is also unable to predict the future effect of the progenies x environments interaction. It was inferred that selection success will always depend on the most accurate possible progeny assessment and using alternatives to reduce the progenies x environments interaction effect.

Determination of breeding strategies for beef cattle bull breeders selling bulls into two competing markets on a non-linear price grid

Computer simulation was used to suggest potential selection strategies for beef cattle breeders with different mixes of clients between two potential markets. The traditional market paid on the basis of carcass weight (CWT), while a new market considered marbling grade in addition to CWT as a basis for payment. Both markets instituted discounts for CWT in excess of 340 kg and light carcasses below 300 kg. Herds were simulated for each price category on the carcass weight grid for the new market. This enabled the establishment of phenotypic relationships among the traits examined [CWT, percent intramuscular fat (IMF), carcass value in the traditional market, carcass value in the new market, and the expected proportion of progeny in elite price cells in the new market pricing grid]. The appropriateness of breeding goals was assessed on the basis of client satisfaction. Satisfaction was determined by the equitable distribution of available stock between markets combined with the assessment of the utility of the animal within the market to which it was assigned. The best goal for breeders with predominantly traditional clients was a CWT in excess of 330 kg, while that for breeders with predominantly new market clients was a CWT of between 310 and 329 kg and with a marbling grade of AAA in the Ontario carcass pricing system. For breeders who wished to satisfy both new and traditional clients, the optimal CWT was 310-329 kg and the optimal marbling grade was AA-AAA. This combination resulted in satisfaction levels of greater than 75% among clients, regardless of the distribution of the clients between the traditional and new marketplaces.

Identification of research to improve the efficiency of breeding strategies for white clover in Australia - a review

A major challenge faced by today's white clover breeder is how to manage resources within a breeding program. It is essential to utilise these resources with sufficient flexibility to build on past progress from conventional breeding strategies, but also take advantage of emerging opportunities from molecular breeding tools such as molecular markers and transformation. It is timely to review white clover breeding strategies. This background can then be used as a foundation for considering how to continue conventional plant improvement activities and complement them with molecular breeding opportunities. In this review, conventional white clover breeding strategies relevant to the Australian dryland target population environments are considered. Attention is given to: (i) availability of genetic variation, (ii) characterisation of germplasm collections, (iii) quantitative models for estimation of heritability, (iv) the role of multi-environment trials to accommodate genotype-by-environment interactions, (v) interdisciplinary research to understand adaptation to dryland environments, (vi) breeding and selection strategies, and (vii) cultivar structure. Current achievements in biotechnology with specific reference to white clover breeding in Australia are considered, and computer modelling of breeding programs is discussed as a useful integrative tool for the joint evaluation of conventional and molecular breeding strategies and optimisation of resource use in breeding programs. Four areas are identified as future research priorities: (i) capturing the potential genetic diversity among introduced accessions and ecotypes that are adapted to key constraints such as summer moisture stress and the use of molecular markers to assess the genetic diversity, (ii) understanding the underlying physiological/morphological root and shoot mechanisms involved in water use efficiency of white clover, with the objective of identifying appropriate selection criteria, (iii) estimation of quantitative genetic parameters of important morphological/physiological attributes to enable prediction of response to selection in target environments, and (iv) modelling white clover breeding strategies to evaluate the opportunities for integration of molecular breeding strategies with conventional breeding programs.

Model selection and error estimation

We study model selection strategies based on penalized empirical loss minimization. We point out a tight relationship between error estimation and data-based complexity penalization: any good error estimate may be converted into a data-based penalty function and the performance of the estimate is governed by the quality of the error estimate. We consider several penalty functions, involving error estimates on independent test data, empirical {\sc vc} dimension, empirical {\sc vc} entropy, andmargin-based quantities. We also consider the maximal difference between the error on the first half of the training data and the second half, and the expected maximal discrepancy, a closely related capacity estimate that can be calculated by Monte Carlo integration. Maximal discrepancy penalty functions are appealing for pattern classification problems, since their computation is equivalent to empirical risk minimization over the training data with some labels flipped.

Selection via simulated individual BLUP based on family genotypic effects in sugarcane

The objective of this work was to propose a new selection strategy for the initial stages of sugarcane improvement, based on the methodology 'simulated individual BLUP (BLUPIS)', which promotes a dynamic allocation of individuals selected in each full-sib family, using BLUP as a base for both the genotypic effects of the referred families and plot effects. The method proposed applies to single full-sib families or those obtained from unbalanced or balanced diallel crosses, half-sib families and self-pollinated families. BLUPIS indicates the number of individuals to be selected within each family, the total number of clones to be advanced, and the number of families to contribute with selected individuals. Correlation between BLUPIS and true BLUP was 0.96, by method validation. Additionally, BLUPIS allows the identification of which replication contains the best individuals of each family.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and deterministic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel metaheuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS metaheuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Feature Selection and Classification Using Age Layered Population Structure Genetic Programming

The curse of dimensionality is a major problem in the fields of machine learning, data mining and knowledge discovery. Exhaustive search for the most optimal subset of relevant features from a high dimensional dataset is NP hard. Sub–optimal population based stochastic algorithms such as GP and GA are good choices for searching through large search spaces, and are usually more feasible than exhaustive and determinis- tic search algorithms. On the other hand, population based stochastic algorithms often suffer from premature convergence on mediocre sub–optimal solutions. The Age Layered Population Structure (ALPS) is a novel meta–heuristic for overcoming the problem of premature convergence in evolutionary algorithms, and for improving search in the fitness landscape. The ALPS paradigm uses an age–measure to control breeding and competition between individuals in the population. This thesis uses a modification of the ALPS GP strategy called Feature Selection ALPS (FSALPS) for feature subset selection and classification of varied supervised learning tasks. FSALPS uses a novel frequency count system to rank features in the GP population based on evolved feature frequencies. The ranked features are translated into probabilities, which are used to control evolutionary processes such as terminal–symbol selection for the construction of GP trees/sub-trees. The FSALPS meta–heuristic continuously refines the feature subset selection process whiles simultaneously evolving efficient classifiers through a non–converging evolutionary process that favors selection of features with high discrimination of class labels. We investigated and compared the performance of canonical GP, ALPS and FSALPS on high–dimensional benchmark classification datasets, including a hyperspectral image. Using Tukey’s HSD ANOVA test at a 95% confidence interval, ALPS and FSALPS dominated canonical GP in evolving smaller but efficient trees with less bloat expressions. FSALPS significantly outperformed canonical GP and ALPS and some reported feature selection strategies in related literature on dimensionality reduction.

Partitions selection strategy for set of clustering solutions

Clustering is a difficult task: there is no single cluster definition and the data can have more than one underlying structure. Pareto-based multi-objective genetic algorithms (e.g., MOCK Multi-Objective Clustering with automatic K-determination and MOCLE-Multi-Objective Clustering Ensemble) were proposed to tackle these problems. However, the output of such algorithms can often contains a high number of partitions, becoming difficult for an expert to manually analyze all of them. In order to deal with this problem, we present two selection strategies, which are based on the corrected Rand, to choose a subset of solutions. To test them, they are applied to the set of solutions produced by MOCK and MOCLE in the context of several datasets. The study was also extended to select a reduced set of partitions from the initial population of MOCLE. These analysis show that both versions of selection strategy proposed are very effective. They can significantly reduce the number of solutions and, at the same time, keep the quality and the diversity of the partitions in the original set of solutions. (C) 2010 Elsevier B.V. All rights reserved.