Model-based Maximum Likelihood Research Articles

Unconstrained feedback controller design using Q-learning from noisy process data

This paper develops a novel model-free Q-learning based approach to estimate linear, unconstrained feedback controllers from noisy process data. The proposed method is based on an extension of an available approach developed to estimate the linear quadratic regulator (LQR) for linear systems with full state measurements driven by Gaussian process noise of known covariance. First, we modify the approach to treat the case of an unknown noise covariance. Then, we use the modified approach to estimate a feedback controller for linear systems with both process and measurement noise and only output measurements. We also present a model-based maximum likelihood estimation (MLE) approach to determine a linear dynamic model and noise covariances from data, which is used to construct a regulator and state estimator for comparisons in simulation studies. The performances of the model-free and model-based controller estimation approaches are compared with an example heating, ventilation, and air-conditioning (HVAC) system. We show that the proposed Q-learning approach estimates a reasonably accurate feedback controller from 24 h of noisy data. The controllers estimated using both the model-free and model-based approaches provide similar closed-loop performances with 3.5 and 2.7% losses respectively, compared to a perfect controller that uses the true dynamic model and noise covariances of the HVAC system. Finally, we give future work directions for the model-free controller design approaches by discussing some remaining advantages of the model-based approaches.

Technical Note: The nearest neighborhood-based approach for estimating basis line-integrals using photon-counting detector.

This study aims to develop a calibration-based estimator for the photon-counting detector (PCD)-based x-ray computed tomography. We propose the nearest neighborhood (NN)-based estimator, which searches for the nearest calibration data for a given PCD output and sets the associated basis line-integrals as the estimate. Searching for the nearest neighbors can be accelerated using the pre-calculated k-d tree for the data. The proposed method is compared to the model-based maximum likelihood (ML) estimator. For slab phantom study, both ML and NN-based methods achieve the Cramér-Rao lower bound and are unbiased for various combinations of three basis materials (water, bone, and gold). The proposed method is also validated for K-edge imaging and presents almost unbiased Au concentrations in the region of interest. The proposed NN-based method is demonstrated to be as accurate as the model-based ML estimator, but it is computationally efficient and requires only calibration measurements.

Genome-wide genetic diversity and population structure of Garcinia kola (Heckel) in Benin using DArT-Seq technology.

Garcinia kola (Heckel) is a versatile tree indigenous to West and Central Africa. All parts of the tree have value in traditional medicine. Natural populations of the species have declined over the years due to overexploitation. Assessment of genetic diversity and population structure of G. kola is important for its management and conservation. The present study investigates the genetic diversity and population structure of G. kola populations in Benin using ultra-high-throughput diversity array technology (DArT) single nucleotide polymorphism (SNP) markers. From the 102 accessions sampled, two were excluded from the final dataset owing to poor genotyping coverage. A total of 43,736 SNPs were reported, of which 12,585 were used for analyses after screening with quality control parameters including Minor allele frequency (≥ 0.05), call rate (≥ 80%), reproducibility (≥ 95%), and polymorphic information content (≥ 1%). Analysis revealed low genetic diversity with expected heterozygosity per population ranging from 0.196 to 0.228. Pairwise F-statistics (FST) revealed low levels of genetic differentiation between populations while an Analysis of molecular variance (AMOVA) indicated that the majority of variation (97.86%) was within populations. Population structure analysis through clustering and discriminant analysis on principal component revealed two admixed clusters, implying little genetic structure. However, the model-based maximum likelihood in Admixture indicated only one genetic cluster. The present study indicated low genetic diversity of G. kola, and interventions are needed to be tailored towards its conservation.

Positively Selected Orthologous Genes Identified in Sesame (Sesamum indicum) by Deep Resequencing

Sesame (Sesamum indicum L.) is the queen of oil seed crops and is cultivated widely in tropical and subtropical areas. The availability of the sesame genome sequence presents unprecedented opportunities for studying its genetics, genomics, and evolution. In this report, we conducted a genome resequencing-based identification of sesame orthologs; in total, 26,379 coding sequences (CDSs) were isolated. Using a reciprocal best hit, we ultimately identified a total of 639 orthologs sets after one-to-one orthologs extraction across seven Pentapetalae plant species. These orthologs were considered to be the most credible between the two species, and in sesame. Furthermore, we performed a branch model-based maximum likelihood estimation of dN/dS of the orthologs, resulting in the identification of 198 evolutionarily accelerated orthologs and 66 positively selected genes (P-value and FDR ＜ 0.05). An enrichment analysis and protein interaction network suggested 19 genes with important functions of the orthologs specific in sesame development and domestication. The method we used here provides a case study for identifying orthologous genes between sesame and other plants species that are distributed in equilibrium phylogenetically, which can be used in other plants.

Drift-Aware Monocular Localization Based on a Pre-Constructed Dense 3D Map in Indoor Environments

Recently, monocular localization has attracted increased attention due to its application to indoor navigation and augmented reality. In this paper, a drift-aware monocular localization system that performs global and local localization is presented based on a pre-constructed dense three-dimensional (3D) map. In global localization, a pixel-distance weighted least squares algorithm is investigated for calculating the absolute scale for the epipolar constraint. To reduce the accumulative errors that are caused by the relative position estimation, a map interaction-based drift detection method is introduced in local localization, and the drift distance is computed by the proposed line model-based maximum likelihood estimation sample consensus (MLESAC) algorithm. The line model contains a fitted line segment and some visual feature points, which are used to seek inliers of the estimated feature points for drift detection. Taking advantage of the drift detection method, the monocular localization system switches between the global and local localization modes, which effectively keeps the position errors within an expected range. The performance of the proposed monocular localization system is evaluated on typical indoor scenes, and experimental results show that compared with the existing localization methods, the accuracy improvement rates of the absolute position estimation and the relative position estimation are at least 30.09% and 65.59%, respectively.

Bayesian spatial analysis and prediction of groundwater contamination in Jhelum city (Pakistan)

Access to clean drinking water, which is essential for healthy human life, is becoming hard with every day, especially in densely populated cities and towns. Ground waters contain many minerals that need to be constantly monitored and in case of any discrepancy with the minerals’ levels immediate remedial measures become necessary to keep it safe to drink. We conduct model-based (aka likelihood based) maximum likelihood estimation and Bayesian kriging predictions for six water quality parameters, namely pH, turbidity, total dissolved solids, calcium, hardness and chloride in groundwater in Jhelum city. Our results show that the concentrations of all the six parameters are, in general, higher in study region, especially four of them have concentrations well above the upper bound of standard limits, of WHO and EPA criteria of USA and Pakistan, and need the immediate attention of concerned authorities for remedial measures.

Regression analysis of current status data in the presence of dependent censoring with applications to tumorigenicity experiments

Current status data frequently occur in many fields including demographic studies and tumorigenicity experiments. In these cases, the censoring or observation time may be correlated to the failure time of interest, the situation that is often referred to as dependent or informative censoring. Although several semiparametric methods have been developed in the literature for the situation, they either only apply to limited situations or may be computationally unstable. To address these, a frailty model-based maximum likelihood approach is proposed with the use of monotone splines to approximate the unknown baseline cumulative hazard function of the failure time. Also a novel EM algorithm, which is based on a three-stage data augmentation and can be easily implemented, is presented. The proposed estimators are proved to be consistent and asymptotically normally distributed. An extensive simulation study is performed to assess the finite sample performance of the proposed approach and suggests that it works well for practical situations. An application to a tumorigenicity study is provided.

Signatures of Dobzhansky-Muller Incompatibilities in the Genomes of Recombinant Inbred Lines.

In the construction of recombinant inbred lines (RILs) from two divergent inbred parents certain genotype (or epigenotype) combinations may be functionally "incompatible" when brought together in the genomes of the progeny, thus resulting in sterility or lower fertility. Natural selection against these epistatic combinations during inbreeding can change haplotype frequencies and distort linkage disequilibrium (LD) relations between loci on the same or on different chromosomes. These LD distortions have received increased experimental attention, because they point to genomic regions that may drive a Dobzhansky-Muller type of reproductive isolation and, ultimately, speciation in the wild. Here we study the selection signatures of two-locus epistatic incompatibility models and quantify their impact on the genetic composition of the genomes of two-way RILs obtained by selfing. We also consider the biases introduced by breeders when trying to counteract the loss of lines by selectively propagating only viable seeds. Building on our theoretical results, we develop model-based maximum-likelihood (ML) tests that can be applied to multilocus RIL genotype data to infer the precise mode of incompatibility as well as the relative fitness of incompatible loci. We illustrate this ML approach in the context of two published Arabidopsis thaliana RIL panels. Our work lays the theoretical foundation for studying more complex systems such as RILs obtained by sibling mating and/or from multiparental crosses.

Treatment effect heterogeneity for univariate subgroups in clinical trials: Shrinkage, standardization, or else.

Treatment effect heterogeneity is a well-recognized phenomenon in randomized controlled clinical trials. In this paper, we discuss subgroup analyses with prespecified subgroups of clinical or biological importance. We explore various alternatives to the naive (the traditional univariate) subgroup analyses to address the issues of multiplicity and confounding. Specifically, we consider a model-based Bayesian shrinkage (Bayes-DS) and a nonparametric, empirical Bayes shrinkage approach (Emp-Bayes) to temper the optimism of traditional univariate subgroup analyses; a standardization approach (standardization) that accounts for correlation between baseline covariates; and a model-based maximum likelihood estimation (MLE) approach. The Bayes-DS and Emp-Bayes methods model the variation in subgroup-specific treatment effect rather than testing the null hypothesis of no difference between subgroups. The standardization approach addresses the issue of confounding in subgroup analyses. The MLE approach is considered only for comparison in simulation studies as the "truth" since the data were generated from the same model. Using the characteristics of a hypothetical large outcome trial, we perform simulation studies and articulate the utilities and potential limitations of these estimators. Simulation results indicate that Bayes-DS and Emp-Bayes can protect against optimism present in the naïve approach. Due to its simplicity, the naïve approach should be the reference for reporting univariate subgroup-specific treatment effect estimates from exploratory subgroup analyses. Standardization, although it tends to have a larger variance, is suggested when it is important to address the confounding of univariate subgroup effects due to correlation between baseline covariates. The Bayes-DS approach is available as an R package (DSBayes).

Tracking crop varieties using genotyping-by-sequencing markers: a case study using cassava (Manihot esculenta Crantz).

BackgroundAccurate identification of crop cultivars is crucial in assessing the impact of crop improvement research outputs. Two commonly used identification approaches, elicitation of variety names from farmer interviews and morphological plant descriptors, have inherent uncertainty levels. Genotyping-by-sequencing (GBS) was used in a case study as an alternative method to track released varieties in farmers’ fields, using cassava, a clonally propagated root crop widely grown in the tropics, and often disseminated through extension services and informal seed systems. A total of 917 accessions collected from 495 farming households across Ghana were genotyped at 56,489 SNP loci along with a “reference library” of 64 accessions of released varieties and popular landraces.ResultsAccurate cultivar identification and ancestry estimation was accomplished through two complementary clustering methods: (i) distance-based hierarchical clustering; and (ii) model-based maximum likelihood admixture analysis. Subsequently, 30 % of the identified accessions from farmers’ fields were matched to specific released varieties represented in the reference library. ADMIXTURE analysis revealed that the optimum number of major varieties was 11 and matched the hierarchical clustering results. The majority of the accessions (69 %) belonged purely to one of the 11 groups, while the remaining accessions showed two or more ancestries. Further analysis using subsets of SNP markers reproduced results obtained from the full-set of markers, suggesting that GBS can be done at higher DNA multiplexing, thereby reducing the costs of variety fingerprinting. A large proportion of discrepancy between genetically unique cultivars as identified by markers and variety names as elicited from farmers were observed. Clustering results from ADMIXTURE analysis was validated using the assumption-free Discriminant Analysis of Principal Components (DAPC) method.ConclusionWe show that genome-wide SNP markers from increasingly affordable GBS methods coupled with complementary cluster analysis is a powerful tool for fine-scale population structure analysis and variety identification. Moreover, the ancestry estimation provides a framework for quantifying the contribution of exotic germplasm or older improved varieties to the genetic background of contemporary improved cultivars.Electronic supplementary materialThe online version of this article (doi:10.1186/s12863-015-0273-1) contains supplementary material, which is available to authorized users.

CNL Disease Resistance Genes in Soybean and Their Evolutionary Divergence.

Disease resistance genes (R-genes) encode proteins involved in detecting pathogen attack and activating downstream defense molecules. Recent availability of soybean genome sequences makes it possible to examine the diversity of gene families including disease-resistant genes. The objectives of this study were to identify coiled-coil NBS-LRR (= CNL) R-genes in soybean, infer their evolutionary relationships, and assess structural as well as functional divergence of the R-genes. Profile hidden Markov models were used for sequence identification and model-based maximum likelihood was used for phylogenetic analysis, and variation in chromosomal positioning, gene clustering, and functional divergence were assessed. We identified 188 soybean CNL genes nested into four clades consistent to their orthologs in Arabidopsis. Gene clustering analysis revealed the presence of 41 gene clusters located on 13 different chromosomes. Analyses of the Ks-values and chromosomal positioning suggest duplication events occurring at varying timescales, and an extrapericentromeric positioning may have facilitated their rapid evolution. Each of the four CNL clades exhibited distinct patterns of gene expression. Phylogenetic analysis further supported the extrapericentromeric positioning effect on the divergence and retention of the CNL genes. The results are important for understanding the diversity and divergence of CNL genes in soybean, which would have implication in soybean crop improvement in future.

Classification of aquatic macrovegetation and substrates with airborne lidar

This study evaluates the potential to use waveform data from bathymetric airborne lidar for mapping seabed substratum and vegetation using a site in the Northern Baltic Sea as a study area. One objective of the study was to develop classification procedures including accurate corrections of waveform data for environmental and lidar system factors, to make them useful for mapping areas with variable water quality. Other objectives were to evaluate the classification accuracy using a combination of depth-derived variables and additional waveform variables, and to compare the results with the classification accuracy obtained when classification is performed using depth-derived variables only (without additional waveform variables). The analysis was based on two waveform variables (bottom pulse width and pulse area), in addition to two depth-derived variables (slope and depth standard deviation). The classification was performed using a model-based maximum likelihood approach. The classification models were created and evaluated with lidar data taken from locations documented with underwater video. The results show that inclusion of the waveform variables significantly improved the classification accuracy. Classification of the seabed into three classes (hard substrate; soft substrate with high vegetation; and soft substrate with low vegetation) had an overall accuracy of 86% when evaluated with an independent data set. This highlights the potential of data from airborne lidar, including waveform data (bottom pulse width and pulse area), for mapping shallow seabed habitats in coastal areas.

Model-Based Bayesian Adaptive Dose-Finding Designs for a Phase II Trial

We describe the planning of a dose-finding study for a compound currently in early Phase II clinical development. Data from a trial with the same primary endpoint are available for a marketed drug that has the same pharmacological mechanism, which provides strong prior information for some characteristics of the new compound. We develop and evaluate informative model-based Bayesian analyses. We also evaluate adaptive designs that change the allocation of patients to doses after an interim analysis. The performance of the model-based Bayesian analyses applied to the adaptive designs is compared to the performance of corresponding analyses applied to a nonadaptive design. The performance is also compared to model-based maximum likelihood estimation and simple pairwise comparison of treatment group means applied to the nonadaptive design to assess the contributions of the model, Bayesian prior distribution, and adaptive designs to improvements in decision and estimation performance.

Bilinear Model-Based Maximum Likelihood Linear Regression Speaker Adaptation Framework

This letter proposes a novel framework for speaker adaptation, using bilinear model-based maximum likelihood linear regression (MLLR) method. First, a set of speaker models is decomposed into the style factor identified as each speaker's characteristics and the common content factor across the speakers, by the bilinear model. Then, using some adaptation data from a new speaker, the speaker-specific model is generated by properly adjusting the dimensionality of the content factor and estimating a new style factor simultaneously. Experimental results show that the proposed framework outperforms MLLR with fewer number of parameters to be estimated.

Molecular characterization of Trichomonas vaginalis isolates from the Philippines

Trichomonas vaginalis is a human urogenital pathogen that causes trichomoniasis, the most common nonviral parasitic sexually transmitted infection in the world. Presently, there are no reports on comparative sequence analysis as well as on the identification of phylogenetic positions of T. vaginalis isolates from the Philippines relative to known trichomonads. In this study, 5.8S rDNA and the flanking internal transcribed spacer (ITS) regions of 57 T. vaginalis isolates were sequenced. The phylogenetic positions of the isolates relative to known trichomonads were determined using the model-based (GTR+Gamma+I) neighbor-joining, maximum likelihood, and Bayesian-inference analyses and the nonmodel-based maximum parsimony analysis. Construction of a phylogenetic tree showed the clustering of all the sequences in one branch together with other T. vaginalis strains obtained through basic local alignment search tool search. Sequencing of the 5.8S rDNA gene and the flanking ITS1and ITS2 regions of T. vaginalis isolates from the Philippines demonstrated low genetic polymorphism. However, comparison of the ribosomal DNA sequences may have implications on some phenotypic characteristics of T. vaginalis.

Topological Estimation Biases with Covarion Evolution

Covarion processes allow changes in evolutionary rates at sites along the branches of a phylogenetic tree. Covarion-like evolution is increasingly recognized as an important mode of protein evolution. Several recent reports suggest that maximum likelihood estimation employing covarion models may support different optimal topologies than estimation using standard rates-across-sites (RAS) models. However, it remains to be demonstrated that ignoring covarion evolution will generally result in topological misestimation. In this study we performed analytical and theoretical studies of limiting distances under the covarion model and four-taxon tree simulations to investigate the extent to which the covarion process impacts on phylogenetic estimation. In particular, we assessed the limits of an RAS model-based maximum likelihood method to recover the phylogenies when the sequence data were simulated under the covarion processes. We find that, when ignored, covarion processes can induce systematic errors in phylogeny reconstruction. Surprisingly, when sequences are evolved under a covarion process but an RAS model is used for estimation, we find that a long branch repel bias occurs.

Molecular Evolution and Phylogeny of the Drosophila saltans Species Group Inferred from the Xdh Gene

The Drosophila saltans group of the subgenus Sophophora consists of five species subgroups whose phylogenetic relationships are poorly known. We have analyzed 2085 coding nucleotides from the xanthine dehydrogenase (Xdh) gene in six species, at least one from each subgroup. We follow a model-based maximum likelihood framework. We first model the substitution process using a tree topology that is approximately accurate. Then we evaluate several candidate tree topologies using a working model of nucleotide substitution. We found that a minimally realistic description of the substitution process along the Xdh region should allow two transition and four transversion rate parameters and different fixed rates for codon positions, which are distributed statistically according to different gamma distributions. The phylogeny obtained using this description differs in significant respects from a phylogeny based on anatomical criteria. We have also analyzed data from five additional (three nuclear and two mitochondrial) gene regions. In our analysis, these relatively short DNA sequences, either separately or jointly, fail to discriminate statistically among alternative phylogenies. When the data for these five gene regions are combined with the Xdh sequences, the strong phylogenetic signal emerging from Xdh becomes somewhat diluted rather than reinforced. The phylogeny of the species and biogeographical considerations suggest that the D. saltans group originated in the tropics of the New World, similarly as the closely related D. willistoni group.

Automatic detection of cyclic alternating pattern (CAP) sequences in sleep: preliminary results

Objectives: The analysis of cyclic alternating pattern (CAP) provides important microstructural information on arousal instability and on EEG synchrony modulation in the sleep process. This work presents a methodology for automatic classification of the micro-organization of human sleep EEG, using the CAP paradigm.Methods: The classification system is composed of 3 parts: feature extraction, detection and classification. The feature extraction part is an EEG generation model-based maximum likelihood estimator. The detector part for the CAP phases A and B is done by a variable length template matched filter, while the classification criteria part is implemented on a state machine ruled-based decision system.Results and conclusions: The preliminary results of the automatic classifier on a group of 4 middle-aged adults are presented. The high agreement between the detector and visual scoring is very promising in the achievement of a fully automated scoring system, although a more exhaustive evaluation program is needed.

Maximum likelihood preprocessing for improved filtered back-projection reconstructions.

Based upon theoretical consideration and experimental investigation, we propose a model-based maximum likelihood (ML) preprocessing approach to the filtered back-projection (FBP) algorithm for application to single photon emission CT (SPECT). In this approach, we optimally removed Poisson noise and an average blur from the projection images and then applied the FBP algorithm for reconstruction. The average blur was due to the combined effect of the geometric response of the imaging system, septal penetration, scatter, and patient motion. We evaluated this preprocessing approach to FBP reconstruction on SPECT physical phantom and patient studies. Preliminary results were encouraging. The advantages of the proposed approach are that statistical noise is accurately modeled, removal of both blur and noise is optimal in the ML sense, and further improvement in this method may be achieved through more accurate modeling of blur in the projection images.

Exact Inference for Contingency Tables with Ordered Categories

Abstract This article proposes an efficient numerical algorithm for small-sample exact inferences in contingency tables having ordinal classifications. The inferences, which apply conditional on the observed marginal totals, also provide an exact analysis for the log-linear model of linear-by-linear association for cell probabilities. An exact test of independence has a one-sided P value equal to the null probability that model-based maximum likelihood estimates of odds ratios are at least as large as the observed estimates. The conditional nonnull distribution yields confidence intervals for odds ratios having a linear-by-linear structure. The computations utilize an extension of the network algorithm proposed by Mehta and Patel (1983).