Inference Research Articles

Variational Label Enhancement.

Multi-label learning focuses on the ambiguity at the label side, i.e., one instance is associated with multiple class labels, where the logical labels are always adopted to partition class labels into relevant labels and irrelevant labels rigidly. However, the relevance or irrelevance of each label corresponding to one instance is essentially relative in real-world tasks and the label distribution is more fine-grained than the logical labels by denoting one instance with a certain number of the description degrees of all class labels. As the label distribution is not explicitly available in most training sets, a process named label enhancement emerges to recover the label distributions in training datasets. By inducing the generative model of the label distribution and adopting the variational inference technique, the approximate posterior density of the label distributions should maximize the variational lower bound. Following the above consideration, LEVI is proposed to recover the label distributions from the training examples. In addition, the multi-label predictive model is induced for multi-label learning by leveraging the recovered label distributions along with a specialized objective function. The recovery experiments on fourteen label distribution datasets and the predictive experiments on fourteen multi-label learning datasets validate the advantage of our approach over the state-of-the-art approaches.

Modelling of South African Hypertension: Comparative Analysis of the Classical and Bayesian Quantile Regression Approaches.

Hypertension has become a major public health challenge and a crucial area of research due to its high prevalence across the world including the sub-Saharan Africa. No previous study in South Africa has investigated the impact of blood pressure risk factors on different specific conditional quantile functions of systolic and diastolic blood pressure using Bayesian quantile regression. Therefore, this study presents a comparative analysis of the classical and Bayesian inference techniques to quantile regression. Both classical and Bayesian inference techniques were demonstrated on a sample of secondary data obtained from South African National Income Dynamics Study (2017–2018). Age, BMI, gender male, cigarette consumption and exercises presented statistically significant associations with both SBP and DBP across all the upper quantiles . The white noise phenomenon was observed on the diagnostic tests of convergence used in the study. Results suggested that the Bayesian approach to quantile regression reveals more precise estimates than the frequentist approach due to narrower width of the 95% credible intervals than the width of the 95% confidence intervals. It is therefore suggested that Bayesian approach to quantile regression modelling to be used to estimate hypertension.

Your Model Trains on My Data? Protecting Intellectual Property of Training Data via Membership Fingerprint Authentication

In recent years, data has become the new oil that fuels various machine learning (ML) applications. Just as the oil refining, providing data to an ML model is a product of massive costs and expertise efforts. However, how to protect the intellectual property (IP) of the training data in ML remains largely open. In this paper, we present MeFA, a novel framework for detecting training data IP embezzlement via Membership Fingerprint Authentication, which is able to determine whether a suspect ML model is trained on the to be protected target data or not. The key observation is that a part of data has a similar influence on the prediction behavior of different ML models. On this basis, MeFA leverages membership inference techniques to extract these data as the fingerprints of the target data and constructs an authentication model to verify the data’s ownership by identifying the obtained membership fingerprints. MeFA has several salient features. It does not assume any knowledge of the suspect model except for its black-box prediction API, through which we can merely get the prediction output of a given input, and also does not require any modification to the dataset or the training process, since it takes advantage of the inherent membership property of the data. As a by-product, MeFA can also serve as a post-protection to verify the ownership of ML models, without modifying the training process of the model. Extensive experiments on three realistic datasets and seven types of ML models validate the effectiveness of MeFA, and demonstrate that it is also robust to scenarios when the training data is partially used or preprocessed with representative membership inference defenses.

Structural Causal Model with Expert Augmented Knowledge to Estimate the Effect of Oxygen Therapy on Mortality in the ICU

Recent advances in causal inference techniques, more specifically, in the theory of structural causal models, provide the framework for identification of causal effects from observational data in the cases where the causal graph is identifiable, i.e., the data generating mechanism can be recovered from the joint distribution. However, no such studies have been done to demonstrate this concept with a clinical example. We present a complete framework to estimate the causal effect from observational data by augmenting expert knowledge in the model development phase and with a practical clinical application. Our clinical application entails a timely and important research question, i.e., the effect of oxygen therapy intervention in the intensive care unit (ICU); the result of this project is useful in a variety of disease conditions, including severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) patients in the ICU. We used data from the MIMIC III database, a standard database in the machine learning community that contains 58,976 admissions from an ICU in Boston, MA, for estimating the oxygen therapy effect on morality. We also identified the covariate-specific effect to oxygen therapy from the model for more personalized intervention.

A Bayesian Inference Technique for Ultrasound Uncertainty Quantification in Metal Additive Manufacturing

A Bayesian Inference Technique for Ultrasound Uncertainty Quantification in Metal Additive Manufacturing

Modelo teórico para levantamento e organização de subsunçores no âmbito da Aprendizagem Significativa

The Meaningful Learning Theory presents two absolutely crucial steps to reach the goal of meaningful teaching. The survey of subsumers, followed by their advanced organization, are conditions of possibility for learning to break the barrier of a merely mechanical learning, aimed at ephemeral memorization. However, such steps are very often disregarded, even by those who explicitly assume this theoretical framework. The survey stage of subsumers, in particular, is riddled with subjectivities on the part of those who organize learning, which brings the extra difficulty of being able to establish comparisons between surveys of subsumers by different teachers. The advanced organization stage, in turn, suggests a personalized teaching approach, since students’ cognitive structures can differ, and usually do, widely. This work seeks to present a methodology that tackles both problems. Without neglecting the inherent subjectivity of the survey of subsumers, it presents a formalization of it, through the fuzzy inference technique, also presenting its computational implementation. The issue of personalization of teaching based on this methodology is also addressed.

A freight origin-destination synthesis model with mode choice

This paper develops a novel procedure to conduct a Freight Origin-Destination Synthesis (FODS) that jointly estimates the trip distribution, mode choice, and the empty trips by truck and rail that provide the best match to the observed freight traffic counts. Four models are integrated: (1) a gravity model for trip distribution, (2) a binary logit model for mode choice, (3) a Noortman and Van Es’ model for truck, and (4) a Noortman and Van Es’ model for rail empty trips. The estimation process entails an iterative minimization of a nonconvex objective function, the summation of squared errors of the estimated truck and rail traffic counts with respect to the five model parameters. Of the two methods tested to address the nonconvexity, an interior point method with a set of random starting points (Multi-Start algorithm) outperformed the Ordinary Least Squared (OLS) inference technique. The potential of this methodology is examined using a hypothetical example of developing a nationwide freight demand model for Bangladesh. This research improves the existing FODS techniques that use readily available secondary data such as traffic counts and link costs, allowing transportation planners to evaluate policy outcomes without needing expensive freight data collection. This paper presents the results, model validation, limitations, and future scope for improvements.

Public debt management and the interaction between fiscal and monetary policies

PurposeThis paper aims to investigate the existing relations between the management of public bonds on the dynamics of debt, term structure of interest rates and economic cycle, through a dynamic stochastic general equilibrium model (DSGE), which was estimated through Bayesian inference techniques using data from Brazil.Design/methodology/approachThe model developed was used to investigate the effects of the public debt average maturity management when the economy faces a monetary policy shock. For this, three management scenarios are evaluated, including Brazilian securities average term.FindingsContrary to what might be inferred from DSGE models that limited the analysis of the debt term by imposing only one-period bonds, a contractionary monetary policy shock does not necessarily cause public debt to increase significantly. Debt term structure plays a crucial role in this result since the government does not need to roll the debt over at higher costs when the debt term profile is longer, reducing the debt service costs and then the impact on the overall debt.Originality/valueDespite the relevance of this theme and its implications for the dynamics of the economy, there is still a gap to be filled in the literature when using DSGE models, since most part of the work that used this methodology limited the analysis of the debt term by imposing that government issues only one-period bonds. This paper differs from the others insofar as it promotes an investigation focused on the role played by debt maturity management on the performance of the contractionary monetary policy. This approach can generate a better understanding of debt management policy and its interaction with fiscal and monetary policies.

Predicting plastic anisotropy using crystal plasticity and Bayesian neural network surrogate models

This work presents an efficient data-driven protocol to accurately predict plastic anisotropy from initial crystallographic texture. In this work, we integrated feed forward neural networks with Variational Bayesian Inference techniques to establish an accurate low-computational cost surrogate model capable of predicting the anisotropic constants based on the texture of the polycrystalline material with quantifiable uncertainty. The developed model was trained on the results of 54,480 crystal plasticity simulations. The performed simulations parametrized Hill’s anisotropic yield model for single crystals and polycrystalline textures, which were robustly represented using generalized spherical harmonics (GSH). Subsequently, the GSH-based representation of the different textures was linked to its corresponding Hill’s anisotropic coefficients using a variational Bayesian neural network. The efficacy and accuracy of the developed surrogate model were critically validated with the results of 20,000 new textures. The predictions from the Bayesian neural network model showed excellent agreement with results obtained from experiments and high-fidelity crystal plasticity finite element simulations.

The U.S. Preventive Services Task Force Abdominal Aortic Aneurysm Screening Guidelines Negligibly Impacted Repair Rates in Male Never-Smokers and Female Smokers

In 2014, in addition to male smokers aged 65-75 years, the U.S. Preventive Services Task Force (USPSTF) recommended abdominal aortic aneurysm (AAA) screening for male never-smokers aged 65-75 years with cardiovascular risk factors (Grade C). The USPSTF evolved from a negative to neutral position on screening for female smokers aged 65-75 years (Grade I). We sought to determine whether 2014 guidelines resulted in more AAA repairs in these populations. We queried the Vascular Quality Initiative national database (2013-2018) for elective endovascular aortic repairs and open aortic repairs. We implemented difference-in-differences (DID) analysis, a causal inference technique that adjusts for secular time trends, to isolate changes in repair numbers due to the 2014 USPSTF guidelines. Our DID models compared changes in repair numbers in patient groups targeted by the USPSTF updates (intervention group) to those in unaffected, older patient groups (control), before and after 2014. The first model compared changes in repair numbers between male never-smokers aged 65-75 years (intervention group) and 76-85 years (control). The second model compared repair numbers between female smokers aged 65-75 years (intervention group) and 76-85 years (control). There was no significant change in male never-smokers (n=1,295) aged 65-75 (42%) vs. 76-85 (58%) undergoing AAA repairs after guideline updates, averaged over 4.5 years (+2.4 percentage points; 95% Confidence Interval [CI] -.56-5.26). However, when their primary insurer was Medicare, male never-smokers aged 65-75 years compared with 76-85 years underwent significantly more repairs over 4.5 years (+3.69 percentage points; 95% CI.16-7.22; representing a 10.4% relative increase from baseline in the proportion of male never-smokers on Medicare undergoing AAA repair). Comparing female smokers (n=2,312) aged 65-75 (54%) vs. 76-85 (46%), there was no significant change in repairs over 4.5 years (-.66 percentage points; 95% CI -4.57-3.26). The USPSTF 2014 AAA guidelines were associated with modestly increased repairs in male never-smokers aged 65-75 years only on Medicare. There was no impact among female smokers. Higher-grade recommendations and improved guideline adherence may be requisites for change.

Accounting for spatial sampling patterns in Bayesian phylogeography

Statistical phylogeography provides useful tools to characterize and quantify the spread of organisms during the course of evolution. Analyzing georeferenced genetic data often relies on the assumption that samples are preferentially collected in densely populated areas of the habitat. Deviation from this assumption negatively impacts the inference of the spatial and demographic dynamics. This issue is pervasive in phylogeography. It affects analyses that approximate the habitat as a set of discrete demes as well as those that treat it as a continuum. The present study introduces a Bayesian modeling approach that explicitly accommodates for spatial sampling strategies. An original inference technique, based on recent advances in statistical computing, is then described that is most suited to modeling data where sequences are preferentially collected at certain locations, independently of the outcome of the evolutionary process. The analysis of georeferenced genetic sequences from the West Nile virus in North America along with simulated data shows how assumptions about spatial sampling may impact our understanding of the forces shaping biodiversity across time and space.

Housing Unit Type and Perceived Social Isolation Among Senior Housing Community Residents.

Social isolation, and its associated health implications, is an important issue for older adults in the United States. To date, there has been limited study of the pathways that connect these 2 factors. The present study expands on previous models by linking factors related to the built environment-in the form of housing unit type-to perceived social isolation among those living independently in dedicated senior housing. The causal inference technique of inverse probability weighting with regression adjustment was employed to assess the impact of living in a townhome-style unit, as opposed to in an apartment building, on self-reported perceived social isolation (N = 1,160). Individuals who lived in townhome-style housing reported a 10.4% lower probability of experiencing social isolation as a result of living in a townhome-style unit as opposed to an apartment building-style unit. The findings provide evidence for the conceptual model that characteristics specific to a given housing unit type may create conditions that exacerbate or buffer individuals from experiencing social isolation. This, in turn, has important implications for the targeting of interventions for social isolation. Policy considerations related to the type of affordable senior housing being built should also be informed by these findings. Additionally, future research should better explicate the role of housing unit type on mental and emotional health outcomes.

Study and Application of Industrial Thermal Comfort Parameters by Using Bayesian Inference Techniques

This paper focuses on the use of Bayesian networks for the industrial thermal comfort issue, specifically in industries in Northern Argentina. Mined data sets that are analyzed and exploited with WEKA and ELVIRA tools are discussed. Thus, networks giving the predictive value of thermal comfort for different pairs of indoor temperature and humidity values according to activity, time, and season, verified in the workplace, were obtained. The results obtained were compared to other statistical models of linear regression used for thermal comfort, thus observing that comfort temperature values are within a same range, yet the network offered more information since a range of options for interior design parameters (temperature/relative humidity) was offered for different work, time, and season conditions. Additionally, if compared with static models of heat exchange, the contribution of Bayesian networks is noted when considering a context of actual operability and adaptability conditions to the environment, which is promising for developing thermal comfort intelligent systems, especially for the development of sustainable settings within the Industry 4.0 paradigm.

Bayesian Inference for Stochastic Cusp Catastrophe Model with Partially Observed Data

The purpose of this paper is to develop a data augmentation technique for statistical inference concerning stochastic cusp catastrophe model subject to missing data and partially observed observations. We propose a Bayesian inference solution that naturally treats missing observations as parameters and we validate this novel approach by conducting a series of Monte Carlo simulation studies assuming the cusp catastrophe model as the underlying model. We demonstrate that this Bayesian data augmentation technique can recover and estimate the underlying parameters from the stochastic cusp catastrophe model.

Technical efficiency of public middle schools in New York City

Using panel data and a “true” fixed effect stochastic frontier model, we estimate persistent and transient technical inefficiency in mathematics (Math) and English Language Arts (ELA) test score gains in New York City public middle schools from 2014 to 2016. We compare several measures of transient technical inefficiency and show that around 58% of New York City middle schools are efficient in Math gains, while 16% are efficient in ELA gains. Multivariate inference techniques are used to determine subsets of efficient schools, providing actionable decision rules to help policymakers target resources and incentives.

Distributed Learning in Wireless Networks: Recent Progress and Future Challenges

The next-generation of wireless networks will enable many machine learning (ML) tools and applications to efficiently analyze various types of data collected by edge devices for inference, autonomy, and decision making purposes. However, due to resource constraints, delay limitations, and privacy challenges, edge devices cannot offload their entire collected datasets to a cloud server for centrally training their ML models or inference purposes. To overcome these challenges, distributed learning and inference techniques have been proposed as a means to enable edge devices to collaboratively train ML models without raw data exchanges, thus reducing the communication overhead and latency as well as improving data privacy. However, deploying distributed learning over wireless networks faces several challenges including the uncertain wireless environment (e.g., dynamic channel and interference), limited wireless resources (e.g., transmit power and radio spectrum), and hardware resources (e.g., computational power). This paper provides a comprehensive study of how distributed learning can be efficiently and effectively deployed over wireless edge networks. We present a detailed overview of several emerging distributed learning paradigms, including federated learning, federated distillation, distributed inference, and multi-agent reinforcement learning. For each learning framework, we first introduce the motivation for deploying it over wireless networks. Then, we present a detailed literature review on the use of communication techniques for its efficient deployment. We then introduce an illustrative example to show how to optimize wireless networks to improve its performance. Finally, we introduce future research opportunities. In a nutshell, this paper provides a holistic set of guidelines on how to deploy a broad range of distributed learning frameworks over real-world wireless communication networks.

Spatial variability characteristics of the effective friction angle of Crag deposits and its effects on slope stability

This study investigated the spatial variability characteristics of the effective friction angle of Crag deposits which are granular soils occur in the east of England. Cone Penetration Test data were obtained at 26 locations and interpreted statistically. The distribution characteristic of the effective friction angle of Crag deposits was derived with the mean value, the standard deviation and the correlation length calibrated. Illustrations were also shown on how factors such as ground water pressures and the existence of soft/organic soil zones affect the measurement of the autocovariance function and thus the correlation length. Bayesian inference technique was adopted alongside the method of moments to determine the correlation length. Based on the obtained statistical parameters, both semi-deterministic (based on standard geotechnical design codes) and probabilistic finite element limit analyses were carried out to investigate the stability of slopes in Crag deposits. Slopes of various inclined angles were considered and comparisons between the semi-deterministic and probabilistic results were conducted to improve the understanding of the stability of Crag slopes and to provide insight into the slope stability code used in practice.

A composite likelihood approach for inference under photometric redshift uncertainty

ABSTRACTObtaining accurately calibrated redshift distributions of photometric samples is one of the great challenges in photometric surveys like LSST, Euclid, HSC, KiDS, and DES. We present an inference methodology that combines the redshift information from the galaxy photometry with constraints from two-point functions, utilizing cross-correlations with spatially overlapping spectroscopic samples, and illustrate the approach on CosmoDC2 simulations. Our likelihood framework is designed to integrate directly into a typical large-scale structure and weak lensing analysis based on two-point functions. We discuss efficient and accurate inference techniques that allow us to scale the method to the large samples of galaxies to be expected in LSST. We consider statistical challenges like the parametrization of redshift systematics, discuss and evaluate techniques to regularize the sample redshift distributions, and investigate techniques that can help to detect and calibrate sources of systematic error using posterior predictive checks. We evaluate and forecast photometric redshift performance using data from the CosmoDC2 simulations, within which we mimic a DESI-like spectroscopic calibration sample for cross-correlations. Using a combination of spatial cross-correlations and photometry, we show that we can provide calibration of the mean of the sample redshift distribution to an accuracy of at least 0.002(1 + z), consistent with the LSST-Y1 science requirements for weak lensing and large-scale structure probes.

Quantitative Analysis of the Impact of Microcredit for Women Microentrepreneurs

Microcredit has a fundamental role as a tool for financial inclusion. It is estimated that there are almost 50 million microentrepreneurs in Brazil, of which two-thirds are women. This work aims to analyze the impact of microcredit for women microentrepreneurs in comparison with men, based on data collected in northeastern Brazil in partnership Avante, a fintech specialized in microcredit. Data were collected from microentrepreneurs who had access to credit (treatment group) and those who did not have their credit granted (control group) so that they could be compared. The analysis was divided into two parts. In Part I, a descriptive statistical analysis of the collected data was performed. In Part II, some simplified causal inference techniques were applied to validate whether the impact of microcredit existed for women microentrepreneurs. The results give strong indications that women, despite having a lower income, grow more than men after access to microcredit. The annualized growth in income of women was 19.87%, while that of men was 14.66%.

Lossless, scalable implicit likelihood inference for cosmological fields

We present a comparison of simulation-based inference to full, field-based analytical inference in cosmological data analysis. To do so, we explore parameter inference for two cases where the information content is calculable analytically: Gaussian random fields whose covariance depends on parameters through the power spectrum; and correlated lognormal fields with cosmological power spectra. We compare two inference techniques: i) explicit field-level inference using the known likelihood and ii) implicit likelihood inference with maximally informative summary statistics compressed via Information Maximising Neural Networks (IMNNs). We find that a) summaries obtained from convolutional neural network compression do not lose information and therefore saturate the known field information content, both for the Gaussian covariance and the lognormal cases, b) simulation-based inference using these maximally informative nonlinear summaries recovers nearly losslessly the exact posteriors of field-level inference, bypassing the need to evaluate expensive likelihoods or invert covariance matrices, and c) even for this simple example, implicit, simulation-based likelihood incurs a much smaller computational cost than inference with an explicit likelihood. This work uses a new IMNN implementation in Jax that can take advantage of fully-differentiable simulation and inference pipeline. We also demonstrate that a single retraining of the IMNN summaries effectively achieves the theoretically maximal information, enhancing the robustness to the choice of fiducial model where the IMNN is trained.