Statistical Learning Strategies Research Articles

Order of statistical learning depends on perceptive uncertainty.

Statistical learning (SL) is an innate mechanism by which the brain automatically encodes the n-th order transition probability (TP) of a sequence and grasps the uncertainty of the TP distribution. Through SL, the brain predicts a subsequent event (e n+1 ) based on the preceding events (e n ) that have a length of "n". It is now known that uncertainty modulates prediction in top-down processing by the human predictive brain. However, the manner in which the human brain modulates the order of SL strategies based on the degree of uncertainty remains an open question. The present study examined how uncertainty modulates the neural effects of SL and whether differences in uncertainty alter the order of SL strategies. It used auditory sequences in which the uncertainty of sequential information is manipulated based on the conditional entropy. Three sequences with different TP ratios of 90:10, 80:20, and 67:33 were prepared as low-, intermediate, and high-uncertainty sequences, respectively (conditional entropy: 0.47, 0.72, and 0.92 bit, respectively). Neural responses were recorded when the participants listened to the three sequences. The results showed that stimuli with lower TPs elicited a stronger neural response than those with higher TPs, as demonstrated by a number of previous studies. Furthermore, we found that participants adopted higher-order SL strategies in the high uncertainty sequence. These results may indicate that the human brain has an ability to flexibly alter the order based on the uncertainty. This uncertainty may be an important factor that determines the order of SL strategies. Particularly, considering that a higher-order SL strategy mathematically allows the reduction of uncertainty in information, we assumed that the brain may take higher-order SL strategies when encountering high uncertain information in order to reduce the uncertainty. The present study may shed new light on understanding individual differences in SL performance across different uncertain situations.

Application of Statistical Learning to Identify Omicron Mutations in SARS-CoV-2 Viral Genome Sequence Data From Populations in Africa and the United States

With timely collection of SARS-CoV-2 viral genome sequences, it is important to apply efficient data analytics to detect emerging variants at the earliest time. To evaluate the application of a statistical learning strategy (SLS) to improve early detection of novel SARS-CoV-2 variants using viral sequence data from global surveillance. This case series applied an SLS to viral genomic sequence data collected from 63 686 individuals in Africa and 531 827 individuals in the United States with SARS-CoV-2. Data were collected from January 1, 2020, to December 28, 2021. The outcome was an indicator of Omicron variant derived from viral sequences. Centering on a temporally collected outcome, the SLS used the generalized additive model to estimate locally averaged Omicron caseload percentages (OCPs) over time to characterize Omicron expansion and to estimate when OCP exceeded 10%, 25%, 50%, and 75% of the caseload. Additionally, an unsupervised learning technique was applied to visualize Omicron expansions, and temporal and spatial distributions of Omicron cases were investigated. In total, there were 2698 cases of Omicron in Africa and 12 141 in the United States. The SLS found that Omicron was detectable in South Africa as early as December 31, 2020. With 10% OCP as a threshold, it may have been possible to declare Omicron a variant of concern as early as November 4, 2021, in South Africa. In the United States, the application of SLS suggested that the first case was detectable on November 21, 2021. The application of SLS demonstrates how the Omicron variant may have emerged and expanded in Africa and the United States. Earlier detection could help the global effort in disease prevention and control. To optimize early detection, efficient data analytics, such as SLS, could assist in the rapid identification of new variants as soon as they emerge, with or without lineages designated, using viral sequence data from global surveillance.

Tracking SARS-CoV-2 Spike Protein Mutations in the United States (January 2020-March 2021) Using a Statistical Learning Strategy.

The emergence and establishment of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants of interest (VOIs) and variants of concern (VOCs) highlight the importance of genomic surveillance. We propose a statistical learning strategy (SLS) for identifying and spatiotemporally tracking potentially relevant Spike protein mutations. We analyzed 167,893 Spike protein sequences from coronavirus disease 2019 (COVID-19) cases in the United States (excluding 21,391 sequences from VOI/VOC strains) deposited at GISAID from 19 January 2020 to 15 March 2021. Alignment against the reference Spike protein sequence led to the identification of viral residue variants (VRVs), i.e., residues harboring a substitution compared to the reference strain. Next, generalized additive models were applied to model VRV temporal dynamics and to identify VRVs with significant and substantial dynamics (false discovery rate q-value < 0.01; maximum VRV proportion >10% on at least one day). Unsupervised learning was then applied to hierarchically organize VRVs by spatiotemporal patterns and identify VRV-haplotypes. Finally, homology modeling was performed to gain insight into the potential impact of VRVs on Spike protein structure. We identified 90 VRVs, 71 of which had not previously been observed in a VOI/VOC, and 35 of which have emerged recently and are durably present. Our analysis identified 17 VRVs ~91 days earlier than their first corresponding VOI/VOC publication. Unsupervised learning revealed eight VRV-haplotypes of four VRVs or more, suggesting two emerging strains (B1.1.222 and B.1.234). Structural modeling supported a potential functional impact of the D1118H and L452R mutations. The SLS approach equally monitors all Spike residues over time, independently of existing phylogenic classifications, and is complementary to existing genomic surveillance methods.

Towards an Integrated PHM framework of the Mutriku Wave Power Plant. Air Turbine Case Study

The Mutriku Wave Power Plant (WPP) is a wave energy conversion plant based on the oscillating water column technology that was commissioned by the Basque Energy Agency in 2011 (Torre-Enciso, Ortubia, De Aguileta, &amp; Marqués, 2009). The Basque Energy Agency is currently the responsible for the operation and maintenance tasks of the plant. From the beginning of the operation of the plant, different degradation and failure events have been reported for WPP components including air-turbines and electric generators (Lekube, Ajuria, Ibeas, Igareta,&amp; Gonzalez,2018), which required unplanned maintenance actions.&#x0D; Despite the thorough monitoring system included in the WPP, a-posteriori root cause failure diagnostics has been challenging due to the lack of experience in similar systems. In this context, without a direct cause-effect correlation between failures and events, the implementation of maintenance actions has been implemented through trial-and-error events, i.e. replacement of components based on intuition and expert knowledge, until the system recovered its optimal operation mode.&#x0D; The Mutriku WPP is designed as a test facility and it is located onshore into the breakwater. Therefore, the operational consequence of unplanned maintenance actions are not as critical as in future commercial open ocean WPPs. However, all the monitored information collected over the years of operation can be used to develop diagnostics models that integrate statistical learning strategies with expert knowledge and accordingly assist engineers in the maintenance decision-making processes of future WPPs.&#x0D; This paper presents an integrated prognostics &amp; health management (PHM) framework for the Mutriku WPP. Figure 1 shows the conceptual block diagram.&#x0D; FIGURE 1 (see attached PDF)&#x0D; The expert knowledge of plant engineers will be combined with collected data and signal-processing methods to detect anomalies, diagnose the failure cause, and predict the remaining useful life (RUL) of plant components and the overall plant (Aizpurua &amp; Catterson, 2015). The development of this approach will permit the prompt detection of anomalies for future operation events and avoid unplanned maintenance actions.&#x0D; The main components evaluated in this paper will be the air-turbines, including different information of the WPP, such as rotational speed, bearing vibration, generated power of the turbine, and pressure loss through the turbine impeller. Firstly, the paper will provide a detailed view of the developed PHM framework for Mutriku WPP (cf. Figure 1). Secondly, after the identification of abnormal patterns, a conditional anomaly detection model will be designed (Catterson, McArthur, &amp; Moss, 2010) from the characteristic operation curve of the turbine and operation condition of the plant as shown in Figure 2.&#x0D; FIGURE 2 (see attached PDF)&#x0D; The proposed approach will be validated with real on-site monitored data. Figure 3 shows the empirical characteristic curve of an air turbine.&#x0D; FIGURE 3 (see attached PDF)&#x0D; Based on the normal operation of the turbines, including contextual information, such as sea-state and plant operational state, probabilistic multivariate models will be developed for the turbines and the operation environment and then their probabilistic correlations will be defined so as to estimate the probability of a turbine being healthy, given the operational information (see Figure 2).&#x0D; Figure 4 shows early results of the anomaly detection model, where it is possible to observe anomalies with very low likelihood.&#x0D; FIGURE 4 (see attached PDF)&#x0D; Vertical dashed line in Figure 4 indicates end of training data, and gray dashed area indicates confidence intervals for improved decision-making under uncertainty.

Neurophysiological Markers of Statistical Learning in Music and Language: Hierarchy, Entropy, and Uncertainty.

Statistical learning (SL) is a method of learning based on the transitional probabilities embedded in sequential phenomena such as music and language. It has been considered an implicit and domain-general mechanism that is innate in the human brain and that functions independently of intention to learn and awareness of what has been learned. SL is an interdisciplinary notion that incorporates information technology, artificial intelligence, musicology, and linguistics, as well as psychology and neuroscience. A body of recent study has suggested that SL can be reflected in neurophysiological responses based on the framework of information theory. This paper reviews a range of work on SL in adults and children that suggests overlapping and independent neural correlations in music and language, and that indicates disability of SL. Furthermore, this article discusses the relationships between the order of transitional probabilities (TPs) (i.e., hierarchy of local statistics) and entropy (i.e., global statistics) regarding SL strategies in human’s brains; claims importance of information-theoretical approaches to understand domain-general, higher-order, and global SL covering both real-world music and language; and proposes promising approaches for the application of therapy and pedagogy from various perspectives of psychology, neuroscience, computational studies, musicology, and linguistics.

Exploration of table and graph literacy of statistics student at Universitas Negeri Makassar

This study aim to explore and review the literacy of tables and graphs for statistics students at Universitas Negeri Makassar. To obtain this aim, descriptive analysis conducted on the students results of table and graph literacy test. This study is importance for leading statistical learning tendency which emphasis on technical procedures with much mathematical formula. From this exploration, we can identify weaknesses that must be looked for solutions, particularly would improve learning of tables and graphs literacy to improve student’s visualization and communication skills. Indicators of tables and graphs literacy have four domains, namely: (1) understanding tables and graphs concepts, (2) the ability of tables and graphs application, (3) skills of making table and graph, and (4) the accuracy of tables and graphs interpretations. The four domains are mapped according to grade of student. Mapping of table and graph literacy will be the basis of statistical learning strategies. Conclusion of the data analysis that the level of achievement scores of students in understanding the concept of the tables and graphs is still low (≤43%), the ability of tables application is still too low (41%), but for the ability of graphics applications is slightly higher (58%). Furthermore, the skills to make the table is still very low (25%), but for the skills to make the graph is slightly higher (59%). The general conclusion that the achievement score of statistics students of students batches 2013, 2014 and 2015 was still below the expectations. Thus, learning strategies need to be formulated to improve every aspect of learning in statistics courses.

Prediction of non-muscle invasive bladder cancer outcomes assessed by innovative multimarker prognostic models.

BackgroundWe adapted Bayesian statistical learning strategies to the prognosis field to investigate if genome-wide common SNP improve the prediction ability of clinico-pathological prognosticators and applied it to non-muscle invasive bladder cancer (NMIBC) patients.MethodsAdapted Bayesian sequential threshold models in combination with LASSO were applied to consider the time-to-event and the censoring nature of data. We studied 822 NMIBC patients followed-up >10 years. The study outcomes were time-to-first-recurrence and time-to-progression. The predictive ability of the models including up to 171,304 SNP and/or 6 clinico-pathological prognosticators was evaluated using AUC-ROC and determination coefficient.ResultsClinico-pathological prognosticators explained a larger proportion of the time-to-first-recurrence (3.1 %) and time-to-progression (5.4 %) phenotypic variances than SNPs (1 and 0.01 %, respectively). Adding SNPs to the clinico-pathological-parameters model slightly improved the prediction of time-to-first-recurrence (up to 4 %). The prediction of time-to-progression using both clinico-pathological prognosticators and SNP did not improve. Heritability (ĥ2) of both outcomes was <1 % in NMIBC.ConclusionsWe adapted a Bayesian statistical learning method to deal with a large number of parameters in prognostic studies. Common SNPs showed a limited role in predicting NMIBC outcomes yielding a very low heritability for both outcomes. We report for the first time a heritability estimate for a disease outcome. Our method can be extended to other disease models.Electronic supplementary materialThe online version of this article (doi:10.1186/s12885-016-2361-7) contains supplementary material, which is available to authorized users.

Co-occurrence statistics as a language-dependent cue for speech segmentation.

To what extent can language acquisition be explained in terms of different associative learning mechanisms? It has been hypothesized that distributional regularities in spoken languages are strong enough to elicit statistical learning about dependencies among speech units. Distributional regularities could be a useful cue for word learning even without rich language-specific knowledge. However, it is not clear how strong and reliable the distributional cues are that humans might use to segment speech. We investigate cross-linguistic viability of different statistical learning strategies by analyzing child-directed speech corpora from nine languages and by modeling possible statistics-based speech segmentations. We show that languages vary as to which statistical segmentation strategies are most successful. The variability of the results can be partially explained by systematic differences between languages, such as rhythmical differences. The results confirm previous findings that different statistical learning strategies are successful in different languages and suggest that infants may have to primarily rely on non-statistical cues when they begin their process of speech segmentation.

A statistical learning strategy for closed-loop control of fluid flows

This work discusses a closed-loop control strategy for complex systems utilizing scarce and streaming data. A discrete embedding space is first built using hash functions applied to the sensor measurements from which a Markov process model is derived, approximating the complex system's dynamics. A control strategy is then learned using reinforcement learning once rewards relevant with respect to the control objective are identified. This method is designed for experimental configurations, requiring no computations nor prior knowledge of the system, and enjoys intrinsic robustness. It is illustrated on two systems: the control of the transitions of a Lorenz 63 dynamical system, and the control of the drag of a cylinder flow. The method is shown to perform well.

Beyond Momentum: Investigating Statistical Learning for Winners--Minus--Losers Anomalies

Financial market anomalies are reviewed and categorized, and the “winners minus losers” anomalies - short-term reversal, Momentum and long-term reversion - are discussed. Literature on extensions to the Momentum anomaly, by incorporating market and fundamental information, is examined. Statistical learning strategies are reviewed, with the goal of applying them to consolidating and improving Momentum.

Patch-primitive driven compressive ghost imaging.

Ghost imaging has rapidly developed for about two decades and attracted wide attention from different research fields. However, the practical applications of ghost imaging are still largely limited, by its low reconstruction quality and large required measurements. Inspired by the fact that the natural image patches usually exhibit simple structures, and these structures share common primitives, we propose a patch-primitive driven reconstruction approach to raise the quality of ghost imaging. Specifically, we resort to a statistical learning strategy by representing each image patch with sparse coefficients upon an over-complete dictionary. The dictionary is composed of various primitives learned from a large number of image patches from a natural image database. By introducing a linear mapping between non-overlapping image patches and the whole image, we incorporate the above local prior into the convex optimization framework of compressive ghost imaging. Experiments demonstrate that our method could obtain better reconstruction from the same amount of measurements, and thus reduce the number of requisite measurements for achieving satisfying imaging quality.

A Flexible PCB Inspection System Based on Statistical Learning

With the large variations in appearance for different kinds of defects in Printed Circuit Boards (PCBs), conventional rule-based inspection algorithms become insufficient for detecting and classifying defects. In this study, an automated PCB inspection system based on statistical learning strategies is developed. First, the partial Hausdorff distance is used to ascertain the positions of defects. Next, the defect patterns are categorized using the Support Vector Machine (SVM) classifier. Defects without regularities in appearance, which cannot be categorized, are identified through the regional defectiveness by comparing the block-wise probability distributions. Experimental results on a real visual inspection platform show that the proposed system is very effective for inspecting a variety of PCB defects.

Learning transcriptional regulatory networks from high throughput gene expression data using continuous three-way mutual information

BackgroundProbability based statistical learning methods such as mutual information and Bayesian networks have emerged as a major category of tools for reverse engineering mechanistic relationships from quantitative biological data. In this work we introduce a new statistical learning strategy, MI3 that addresses three common issues in previous methods simultaneously: (1) handling of continuous variables, (2) detection of more complex three-way relationships and (3) better differentiation of causal versus confounding relationships. With these improvements, we provide a more realistic representation of the underlying biological system.ResultsWe test the MI3 algorithm using both synthetic and experimental data. In the synthetic data experiment, MI3 achieved an absolute sensitivity/precision of 0.77/0.83 and a relative sensitivity/precision both of 0.99. In addition, MI3 significantly outperformed the control methods, including Bayesian networks, classical two-way mutual information and a discrete version of MI3. We then used MI3 and control methods to infer a regulatory network centered at the MYC transcription factor from a published microarray dataset. Models selected by MI3 were numerically and biologically distinct from those selected by control methods. Unlike control methods, MI3 effectively differentiated true causal models from confounding models. MI3 recovered major MYC cofactors, and revealed major mechanisms involved in MYC dependent transcriptional regulation, which are strongly supported by literature. The MI3 network showed that limited sets of regulatory mechanisms are employed repeatedly to control the expression of large number of genes.ConclusionOverall, our work demonstrates that MI3 outperforms the frequently used control methods, and provides a powerful method for inferring mechanistic relationships underlying biological and other complex systems. The MI3 method is implemented in R in the "mi3" package, available under the GNU GPL from and from the R package archive CRAN.

Statistical learning applied to computer-assisted fish age and growth estimation from otolith images

Computer-assisted tools need to be developed to help in the accurate and efficient acquisition of fish age and growth data for ecological and assessment issues. Stating fish age and growth analysis as pattern classification issues, the proposed approach relies on a statistical learning strategy. Given otolith images interpreted by an expert, probabilistic kernel-based methods (namely Kernel Logistic Regression) are used to infer interpretation rules. More precisely, two different probabilistic models are introduced: one to infer fish age from otolith images and a second one aiming at evaluating whether or not a given otolith growth pattern is realistic w.r.t. training examples. These probabilistic models provide us with the basis for coping with three different issues: the automated estimation of fish age from otolith images, the estimation of individual otolith growth patterns, and the definition of a confidence measure of otolith interpretations. These computer-assisted ageing tools are validated for a dataset of plaice otoliths.

Automated fish age estimation from otolith images using statistical learning

Abstract The acquisition of age and growth data is of key importance for fisheries research (assessment, marine ecology issues, etc.). Consequently, automating this task is of great interest. In this paper, we investigate the use of statistical learning techniques for fish age estimation. The core of this study lies in the definition of relevant image-related features. We rely on the computation of a 1D representation summing up the content of otolith images within a predefined area of interest. Features are then extracted from this non-stationary representation depicting the alternation of seasonal growth rings. Thus, fish age estimation can be viewed as a multi-class classification issue using statistical learning strategies. In particular, a procedure based on demodulation and remodulation of fish growth patterns is used to improve the generalization properties of the trained classifiers. The experimental evaluation is carried out over a dataset of 320 plaice otolith images from age groups 1–6. We analyze both, the performances of several statistical classifiers, namely SVMs (support vector machines) and neural networks, and the relevance of the proposed image-based feature sets. In addition, the combination of additional biological and shape features to the image-related ones is considered. We reach a rate of correct age estimation of 88% w.r.t. the expert ground truth. This demonstrates the relevance of the proposed approach for the automation of routine aging and for computer-assisted aging.