Likelihood Assignment Research Articles

Understanding Likelihood of Normalizing Flow and Image Complexity through the Lens of Out-of-Distribution Detection

Out-of-distribution (OOD) detection is crucial to safety-critical machine learning applications and has been extensively studied. While recent studies have predominantly focused on classifier-based methods, research on deep generative model (DGM)-based methods have lagged relatively. This disparity may be attributed to a perplexing phenomenon: DGMs often assign higher likelihoods to unknown OOD inputs than to their known training data. This paper focuses on explaining the underlying mechanism of this phenomenon. We propose a hypothesis that less complex images concentrate in high-density regions in the latent space, resulting in a higher likelihood assignment in the Normalizing Flow (NF). We experimentally demonstrate its validity for five NF architectures, concluding that their likelihood is untrustworthy. Additionally, we show that this problem can be alleviated by treating image complexity as an independent variable. Finally, we provide evidence of the potential applicability of our hypothesis in another DGM, PixelCNN++.

The AILA Methodology for Automated and Intelligent Likelihood Assignment in Risk Assessment

This article recognises the widespread application of risk assessment in ICT and aims at reducing the influence of human subjectivity and distraction by means of a methodology for the Automated and Intelligent Likelihood Assignment (AILA). The AILA Methodology, with its various components, applies when risk assessment proceeds exclusively upon information stated in a policy coming as a text document. This scenario is extremely common through small to medium sized institutions. Among the main contributions of this article lies the AILA Entity Extractor, which facilitates the risk assessor in the identification of entities, then of assets, from a given policy. Then, the AILA Classifier automates the assignment of likelihood values to given threats for assets. Moreover, the synergy of AILA with an existing tool for risk assessment demonstrates how to achieve more objective likelihood assignments. AILA is general in support of any risk assessment and, for the sake of demonstration, is applied to assess the privacy risk induced over physical persons by three real-world manufacturers from the automotive domain, namely Toyota, Mercedes and Tesla. AILA is also validated against a risk assessment methodology by ENISA, thereby confirming effectiveness and efficiency of the new methodology (which is dramatically more automated than ENISA’s). AILA combines and consolidates together several techniques in an unprecedented fashion, including Natural Language Processing by summarisation and entity recognition, dataset labelling by appeal to the ToS;DR service, and fully-supervised Machine Learning and regression analysis. Finally, to contribute to open knowledge, the general, executable components of AILA, the AILA Entity Extractor and the AILA Classifier are released open source along with the the privacy-specific components, the AILA Privacy Dataset and the AILA Privacy Model.

Childhood and Adulthood Residency Prediction Using Strontium (87Sr/86Sr) and Oxygen (δ18O) Isotopes for Unidentified Deceased Migrants Recovered in Southern Texas

Strontium and oxygen isotope values are used in conjunction with a maximum likelihood assignment model to predict the childhood and adulthood geographic residences of 20 deceased migrants recovered along the México-U.S. border in south Texas. The objectives of this research are to determine (1) if the childhood and adulthood residency assignments predict the same geographic region, (2) if using the predicted residential history/nationality as a filter on the National Missing and Unidentified Persons System (NamUs) reduces the number of potential matching missing person reports, and (3) if we should target teeth or bone as the optimal hard tissue to sample. We found that 12 of the 20 cases (60%) predict similar regions of residency using isotopes derived from bone bioapatite (adulthood residency) and tooth bioapatite (childhood residency). When predicted residential history/nationality data were used as a filter in NamUs, there was a significant reduction in the mean number of case searches (mean = 1568.5 ± 505.3 [1 SD] vs. 107.9 ± 142.6 [1 SD]; t = 12.441, df = 22, p < 0.001, unequal variances assumed). These results demonstrate the utility of isotope data for narrowing down the region of origin and in turn the number of missing persons records to review. Finally, we recommend that death investigators at minimum collect dental samples (premolars or molars) for future unidentified migrant cases to aid in identification efforts utilizing isotope analytical methods. Bone bioapatite samples can provide more recent residential history information but should be used in conjunction with tooth enamel data.

Reproducibility of aortic valve calcification scoring with computed tomography – An interplatform analysis

BackgroundTo investigate whether aortic valve calcification (AVC) scoring performed with different workstation platforms generates comparable and thus software-independent results. MethodsIn this IRB-approved retrospective study, we included 100 consecutive patients with symptomatic aortic stenosis undergoing CT prior to transcatheter aortic valve implantation. Two independent observers performed AVC scoring on non-enhanced images with commercially available software platforms of four vendors (GE, Philips, Siemens, 3mensio). Gender-specific Agatston score cut-off values were applied according to current recommendations to assign patients to different likelihood categories of aortic stenosis (unlikely to very likely). Comparative analysis of Agatston scores between the four platforms were performed by using Kruskal-Wallis analysis, Spearman rank correlation, linear regression analysis, and Bland-Altman analysis. Differences in category assignment were compared using Fisher's exact test and Cohen's kappa. ResultsFor both observers, each workstation platform produced slightly different numeric AVC Agatston scores, however, without statistical significance (p = 0.96 and p = 0.98). Excellent correlation was found between platforms, with r = 0.991–0.996 (Spearman) and r2 = 0.981–0.992 (regression analysis) for both observers. Bland-Altman analyses revealed small mean differences with narrow limits of agreement between platforms (mean differences: 6 ± 128 to 100 ± 179), for inter-observer (mean differences: 1 ± 43 to 12 ± 70), and intra-observer variability (mean differences: 9 ± 42 to 20 ± 96). Observer 1 assigned 11 (kappa: 0.85–0.97) and observer 2 assigned 10 patients (kappa: 0.88–0.95) to different likelihood groups of severe aortic stenosis with at least one platform. Overall, there was no significant difference of likelihood assignment between platforms (p = 0.98 and p = 1.0, respectively). ConclusionWhile absolute values differ slightly, common commercially available software platforms produce comparable results for AVC scoring, which indicates software-independence of the method.

Genetic Diversity of Oilseed Rape Fields and Feral Populations in the Context of Coexistence with GM Crops.

Despite growing concern about transgenes escaping from fields, few studies have analysed the genetic diversity of crops in an agroecosystem over several years. Accurate information about the dynamics and relationship of the genetic diversity of crops in an agroecosystem is essential for risk assessment and policies concerning the containment of genetically modified crops and their coexistence with crops grown by conventional practices. Here, we analysed the genetic diversity of oilseed rape plants from fields and feral populations over 4 years in an agricultural landscape of 41 km2. We used exact compatibility and maximum likelihood assignment methods to assign these plants to cultivars. Even pure lines and hybrid cultivar seed lots contained several genotypes. The cultivar diversity in fields reflected the conventional view of agroecosystems quite well: that is, there was a succession of cultivars, some grown for longer than others because of their good performance, some used for one year and then abandoned, and others gradually adopted. Three types of field emerged: fields sown with a single cultivar, fields sown with two cultivars, and unassigned fields (too many cultivars or unassigned plants to reliably assign the field). Field plant diversity was higher than expected, indicating the persistence of cultivars that were grown for only one year. The cultivar composition of feral populations was similar to that of field plants, with an increasing number of cultivars each year. By using genetic tools, we found a link between the cultivars of field plants in a particular year and the cultivars of feral population plants in the following year. Feral populations on road verges were more diverse than those on path verges. All of these findings are discussed in terms of their consequences in the context of coexistence with genetically modified crops.

Fragment assignment in the cloud with eXpress-D

BackgroundProbabilistic assignment of ambiguously mapped fragments produced by high-throughput sequencing experiments has been demonstrated to greatly improve accuracy in the analysis of RNA-Seq and ChIP-Seq, and is an essential step in many other sequence census experiments. A maximum likelihood method using the expectation-maximization (EM) algorithm for optimization is commonly used to solve this problem. However, batch EM-based approaches do not scale well with the size of sequencing datasets, which have been increasing dramatically over the past few years. Thus, current approaches to fragment assignment rely on heuristics or approximations for tractability.ResultsWe present an implementation of a distributed EM solution to the fragment assignment problem using Spark, a data analytics framework that can scale by leveraging compute clusters within datacenters–“the cloud”. We demonstrate that our implementation easily scales to billions of sequenced fragments, while providing the exact maximum likelihood assignment of ambiguous fragments. The accuracy of the method is shown to be an improvement over the most widely used tools available and can be run in a constant amount of time when cluster resources are scaled linearly with the amount of input data.ConclusionsThe cloud offers one solution for the difficulties faced in the analysis of massive high-thoughput sequencing data, which continue to grow rapidly. Researchers in bioinformatics must follow developments in distributed systems–such as new frameworks like Spark–for ways to port existing methods to the cloud and help them scale to the datasets of the future. Our software, eXpress-D, is freely available at: http://github.com/adarob/express-d.

OLD AGE SEX: A PARENTAGE STUDY OF DIFFERENT AGE COHORTS IN A NATIVE VETERAN PEDUNCULATE OAK (QUERCUS ROBUR L.) WOODLAND USING MICROSATELLITE MARKERS

To quantify levels of gene flow among and genetic variation within different age cohorts of pedunculate oak (Quercus robur L.) in a semi-natural oak woodland, diversity and parentage studies were undertaken on a stand in Ballytobin, Co. Kilkenny, Ireland. Trees were categorised into four size classes—seedlings, saplings, mature trees and veterans, and genetic differentiation was estimated for three microsatellite markers using AMOVA. Parentage was assessed using a likelihood assignment approach. Between nine and fourteen alleles were detected per locus (mean gene diversity, h, 0.8312). All samples from Ballytobin were heterozygous. AMOVA between age cohorts (veterans, mature, saplings and seedlings) estimated that only 2% of the variation was partitioned among age cohort and the mean FST was low (0.015) but significant (P = 0.034). Parent-pair analysis revealed that out of 24 candidate trees, fifteen were assigned as parents to at least one offspring and, out of these, five were veteran. The average pollination distance was estimated at 47.8m with seven out of the twelve parent-pairs being within 50m of each other. The measured range of distances between trees indicated in the analysis as parents was between 22m and 150m. The highest values for number of parent/offspring assignments were between 40m and 49m. Overall, veteran oak at the site were found to be reproductively successful and a low level of differentiation was found between age cohorts indicating that population genetic diversity has been maintained over time.

Old Age Sex: A Parentage Study of Different Age Cohorts in a Native Veteran Pedunculate Oak (<i>Quercus Robur L.</i>) Woodland Using Microsatellite Markers

To quantify levels of gene flow among and genetic variation within different age cohorts of pedunculate oak ( Quercus robur L.) in a semi-natural oak woodland, diversity and parentage studies were undertaken on a stand in Ballytobin, Co. Kilkenny, Ireland. Trees were categorised into four size classes seedlings, saplings, mature trees and veterans, and genetic differentiation was estimated for three microsatellite markers using AMOVA. Parentage was assessed using a likelihood assignment approach. Between nine and fourteen alleles were detected per locus (mean gene diversity, h, 0.8312). All samples from Ballytobin were heterozygous. AMOVA between age cohorts (veterans, mature, saplings and seedlings) estimated that only 2% of the variation was partitioned among age cohort and the mean FSt was low (0.015) but significant (P = 0.034). Parent-pair analysis revealed that out of 24 candidate trees, fifteen were assigned as parents to at least one offspring and, out of these, five were veteran. The average pollination distance was estimated at 47.8m with seven out of the twelve parent-pairs being within 50m of each other. The measured range of distances between trees indicated in the analysis as parents was between 22m and 150m. The highest values for number of parent/ offspring assignments were between 40m and 49m. Overall, veteran oak at the site were found to be reproductively successful and a low level of differentiation was found between age cohorts indicating that population genetic diversity has been maintained over time.

Possible cryptic stock structure for minke whales in the North Atlantic: Implications for conservation and management

The minke whale is the last of the great whale species to be hunted in significant numbers. Effective management must include an understanding of how genetic diversity is divided and distributed among putative local populations, and as for many migratory species, this is complicated for the minke whale by large-scale seasonal movement among geographic regions. The problem is that the geographic identity of breeding populations is not known, and instead these whales are predictably found and hunted where different breeding stocks may mix on seasonal feeding grounds. Here we use microsatellite DNA and mtDNA markers to investigate minke whale population structure across the species’ range in the North Atlantic. We found no evidence of geographic structure comparing putative populations in recognized management areas, though some limited structure had been indicated in earlier studies. However, using individual genotypes and likelihood assignment methods, we identified two putative cryptic stocks distributed across the North Atlantic in similar proportions in different regions. Some differences in the proportional representation of these populations may explain some of the apparent differentiation between regions detected previously. The implication would be that minke whales range extensively across the North Atlantic seasonally, but segregate to some extent on at least two breeding grounds. This means that established stock boundaries in the North Atlantic, currently used for management, should be re-considered to ensure the effective conservation of genetic diversity.

THE BREED TRACEABILITY OF SHEEP MEAT BY USING MOLECULAR GENETICS METHODS: PRELIMINARY RESULTS

Safety and quality foods of animal origin are extremely important for consumers. The aim of this work was to evaluate the feasibility of a method to track the breed origin of sheep meat all along the production chain using molecular genetics tools. A total of 800 samples evenly distributed among seven Italian sheep breeds have been typed at 19 STR markers, together with 90 samples from both imported sheep animals and local crossbred animals withdrawn at slaughterhouses. A maximum likelihood assignment test was adopted to evaluate STR ability to allocate samples to their true breed of origin. Sarda animals were all correctly allocated, as well as more than 98% of samples from the other breeds. Only slightly worst allocation performances were observed for imported and crossbred animals. Preliminary results seem quite promising, though further analyses will be needed in order to better understand the statistical power of such an assignment test before implementation in the sheep meat production chain.

CRANIAL DESCRIPTION AND GENETIC IDENTITY OF THE HOLOTYPE SPECIMEN OF TURSIOPS ADUNCUS (EHRENBERG, 1832)

AbstractTwo species of bottlenose dolphins are currently recognized by most cetologists: the pan‐tropical and temperate Tursiops truncatus and the endemic Indo‐Pacific T. aduncus. The latter was described from a specimen from the Red Sea, with nothing in the description that would allow referral of the specimen to one or the other of the two species. Because both species occur in the northern Indian Ocean, it was possible that the holotype specimen was actually a common bottlenose dolphin, not of the Indo‐Pacific species. The holotype skull was thought lost but has been found in the Berlin Museum. We describe the skull and examine its affinities through comparison of a partial mtDNA control‐region sequence with sequences from South Africa and through application of classification functions from a discriminant analysis of the two putative species from Taiwan. The mtDNA sequence is identical to that of South African specimens referred to the Indo‐Pacific species, and the multivariate likelihood assignment associates the skull with Taiwanese specimens referred to that species. These results ensure the correctness of use of the name T. aduncus for the species.

THE ROLE OF SPECIES ABUNDANCE IN DETERMINING BREEDING ORIGINS OF MIGRATORY BIRDS WITH STABLE ISOTOPES

Understanding the breeding origins of migratory birds captured on their wintering grounds has important management and conservation implications for declining populations of songbirds. Stable isotopes have recently been used to infer origins for species where application of conventional markers fails. A natural method of linking wintering birds to breeding populations, and one that has not been previously applied to stable isotopes, is based on likelihood. Using a likelihood assignment rule, birds are associated with the breeding population under which their realized isotope signature is most likely to have been generated. We report the first illustration of using likelihood-based assignment with stable isotope data. Moreover, within a probability-based framework for assignment, we argue that a more natural formulation of the assignment problem should be based on the probability of origin given the observed data, or the posterior probability of origin. The relationship between posterior assignment and that based on simple likelihood is embodied in Bayes Rule, which establishes a clear linkage between the distribution of the breeding population (i.e., relative abundance) and probability of origin. We demonstrate likelihood and posterior assignment using a large data set on Black-throated Blue Warblers (Dendroica caerulescens). Our results suggest that relative abundance is likely to be a more crucial consideration in the presence of less geographically structured isotopes, when the distribution of abundance is highly non-uniform, or when the range of the species is geographically restricted.

Microsatellite DNA variation in Atlantic sturgeon (Acipenser oxyrinchus oxyrinchus) and cross-species amplification in the Acipenseridae

Overharvest and habitat alteration have led toa collapse of most commercial Atlantic sturgeon(Acipenser oxyrinchus oxyrinchus)fisheries while pushing the species to rarityor extirpation in most of its historical range. A biologically sound conservation program forthis species requires knowledge of its geneticdiversity and of the evolutionary relationshipsamong geographic populations. To address theseresearch needs, six microsatellite loci wereisolated from A. o. oxyrinchus. Pedigreeanalysis suggested that all are inherited in acodominant Mendelian pattern. The six lociwere tested in ten additional sturgeon speciesfrom three genera and three apparent ploidylevels (4n, 8n, 16n). Approximately 70% ofsuccessful locus-species amplifications werepolymorphic. Polysomy was observed most oftenin 8n and 16n species. Genetic diversity andpopulation structure of A. o. oxyrinchuswere assayed using three polymorphic Aoxmarkers and four markers developed from lakesturgeon (A. fulvescens). A. o.oxyrinchus were sampled from the AltamahaRiver, Georgia, USA north to the St. LawrenceRiver, Quebec, Canada. Gulf sturgeon, A.o. desotoi, were sampled from the SuwanneeRiver, Florida, USA, to assess differentiationbetween the subspecies. Seventy-seven alleleswere observed to segregate into uniquemultilocus genotypes for each of the 392individuals assayed. Mean diversity wasgreatest in the Chesapeake Bay (9.7 alleles perlocus) and Delaware River (7.4 alleles perlocus) collections, and lowest in the St.Lawrence River (4.6 alleles per locus). Meanheterozygosity across seven loci ranged from44.3% (St. Lawrence River) to 62.6% (Altamaha River). Significant allelic heterogeneity wasobserved in 82% of pairwise comparisons aswell as a global test (p < 0.0001) for A.o. oxyrinchus collections. Genetic distancesuggests the presence of at least sixsubpopulations in A. o. oxyrinchus: St.Lawrence River, St. John River, Hudson River,Delaware River, Albemarle Sound, and AltamahaRiver. Genetic and geographic distances werepositively correlated (r = 0.57, p < 0.03) amongA. o. oxyrinchus, suggesting isolation bydistance and philopatry. Hierarchical genediversity analysis indicated significantgenetic population structure at every level. Maximum likelihood assignment tests correctlyassigned individual fish to collection with ahigh rate of success (mean = 87.5%); this andother lines of evidence indicated that theChesapeake Bay collection represents a mixedpopulation of sub-adult sturgeon from northernand southern Atlantic coast populations. Population structure was correlated with thatsuggested by earlier mitochondrial (mt) DNAanalyses. Significant diversity was observedbetween two Canadian populations from whichonly a single mtDNA haplotype has beenreported.

Use of error matrices to improve area estimates with maximum likelihood classification procedures

The maximum likelihood classifier is by far the most widespread among supervised classification methods. This procedure offers numerous advantages, but it has considerable shortcomings in the presence of strongly irregular spectral distributions, mainly related to bias in area estimates. Since these cases are quite common, some methods have already been proposed to correct biased area estimates from maximum likelihood classifications, but they are often not generally applicable or statistically stable. In this article a method is put forward to correct maximum likelihood assignment probabilities by means of a transition matrix; this matrix is derived through a simple mathematical algorithm from a contingency table of a previous classification compared to reference pixels. The purpose is clearly to attain a diagonalization of the final error sources to better estimate area extents and, above all, to achieve higher global discrimination accuracy. As different environmental situations may cause wide variability in the results of such a procedure, this was tested in three case studies using TM data acquired over areas with different landscapes. The results, evaluated by means of suitable statistics, significantly support that the method has general validity and applicability.

A circular median filter approach for resolving directional ambiguities in wind fields retrieved from spaceborne scatterometer data

A common problem associated with spaceborne scatterometry, including SASS, NSCAT and ERS‐1, is that retrieval algorithms often compute multiple wind vectors per retrieval cell. Only one wind vector per retrieval cell corresponds to the true solution while the others are aliases that arise from symmetries in the response of small scale surface waves to wind forcing. The NSCAT retrieval algorithm ranks each ambiguous wind vector within a retrieval cell according to the likelihood that the wind vector is the true solution. Although the most likely wind vectors are usually correct, in approximately 40% of the retrieval cells an alias is ranked first. Based on the mathematical properties of these likelihood assignment errors, an ambiguity removal algorithm is derived that utilizes a nonlinear circular median filter (CMF) to select the true solution in each retrieval cell. The CMF algorithm was tested by analyzing twelve simulated wind fields with an average likelihood assignment error rate of 389 per 1000 retrievals. After processing by the CMF algorithm, the error rate was significantly reduced to an average of only 36 per 1000. To better understand the performance of the CMF algorithm under operational conditions, the analysis was extended by adding spatially correlated likelihood assignment errors to one of the simulated wind fields. A re‐analysis of this wind field showed that the performance of the CMF algorithm may decrease significantly if the likelihood assignment errors are correlated over several retrieval cells. In addition to deriving an ambiguity removal algorithm, a wind field smoothing technique that utilizes a circular median filter is derived and tested on one of the resolved wind fields. The results showed that CMF smoothing technique increases the spatial coherency between the true and resolved wind fields. Based on these analyses, specific recommendations are made for monitoring the performance of the wind retrieval process and for improving the accuracy of the retrieved wind fields.