Learning Literature Research Articles

The relative data hungriness of unpenalized and penalized logistic regression and ensemble-based machine learning methods: the case of calibration.

Machine learning methods are increasingly being used to predict clinical outcomes. Optimism is the difference in model performance between derivation and validation samples. The term "data hungriness" refers to the sample size needed for a modelling technique to generate a prediction model with minimal optimism. Our objective was to compare the relative data hungriness of different statistical and machine learning methods when assessed using calibration. We used Monte Carlo simulations to assess the effect of number of events per variable (EPV) on the optimism of six learning methods when assessing model calibration: unpenalized logistic regression, ridge regression, lasso regression, bagged classification trees, random forests, and stochastic gradient boosting machines using trees as the base learners. We performed simulations in two large cardiovascular datasets each of which comprised an independent derivation and validation sample: patients hospitalized with acute myocardial infarction and patients hospitalized with heart failure. We used six data-generating processes, each based on one of the six learning methods. We allowed the sample sizes to be such that the number of EPV ranged from 10 to 200 in increments of 10. We applied six prediction methods in each of the simulated derivation samples and evaluated calibration in the simulated validation samples using the integrated calibration index, the calibration intercept, and the calibration slope. We also examined Nagelkerke's R2, the scaled Brier score, and the c-statistic. Across all 12 scenarios (2 diseases × 6 data-generating processes), penalized logistic regression displayed very low optimism even when the number of EPV was very low. Random forests and bagged trees tended to be the most data hungry and displayed the greatest optimism. When assessed using calibration, penalized logistic regression was substantially less data hungry than methods from the machine learning literature.

Shifting Under Pressure: Timing Practices within Regulatory Information Disclosure Programs

Research examines the process through which regulatory agencies and other powerful third parties publicly disseminate information about organizations in order to reduce the information asymmetry between those organizations and their dispersed or less powerful stakeholders. These disclosure programs are presumed to work by mobilizing stakeholders to pressure organizations for change. Prior studies focus on organizations’ resistance to that external pressure, often drawing from theories of institutional forces and organizational legitimacy. Instead, the present study examines the public disclosure process itself, and adapts research on organizational learning and collective attention focus to develop new theory regarding how organizations may selectively time their practices to influence these disclosures in their favor. Results suggest that organizations may respond to disclosure programs and other governance structures when actively monitored, but may drift out of compliance during intervening periods. Findings also suggest that strong governance outside of the regulatory disclosure process can help to reduce these problems. The study contributes to corporate governance, organizational learning, and other literatures by developing new theory regarding the timing of organizational responses to regulatory interventions, and by providing evidence that overlapping forms of oversight may mitigate the limitations within any one particular governance mechanism.

Modeling noisy learning in a dynamic oligopoly experiment

Estimating demand before production poses a significant challenge for many industries, including vaccine manufacturing, newspapers, and perishable foods. Both industry professionals and experimental subjects in laboratory settings often struggle to determine optimal production. This paper sheds light on the cognitive processes that explain individuals' inability to act optimally, using data from Nagel and Vriend (1999a,b). In their experiment, participants, acting as firms, set production levels without prior knowledge of the demand generated by own and competitors' advertising efforts. We first reexamine their two-step learning model, which involves directional learning for production levels and a hill-climbing algorithm for advertising, utilizing exogenous adjustment size distributions. We improve upon this model, inspired by the macroeconomic learning literature, with agents using constant gain learning for production decisions and hill climbing for advertising decisions with endogenous adjustments. We demonstrate that our model provides a better fit to the data than the original model by Nagel and Vriend (1999a) and yields superior out-of-sample forecasts, both for one-period-ahead predictions and simulated paths.

Reporting bias, not external focus: A robust Bayesian meta-analysis and systematic review of the external focus of attention literature.

Evidence has ostensibly been accumulating over the past 2 decades suggesting that an external focus on the intended movement effect (e.g., on the golf club during a swing) is superior to an internal focus on body movements (e.g., on your arms during a swing) for skill acquisition. Seven previous meta-studies have all reported evidence of external focus superiority. The most comprehensive of these concluded that an external focus enhances motor skill retention, transfer, and performance and leads to reduced eletromyographic activity during performance and that more distal external foci are superior to proximal external foci for performance. Here, we reanalyzed these data using robust Bayesian meta-analyses that included several plausible models of publication bias. We found moderate to strong evidence of publication bias for all analyses. After correcting for publication bias, estimated mean effects were negligible: g = 0.01 (performance), g = 0.15 (retention), g = 0.09 (transfer), g = 0.06 (electromyography), and g = -0.01 (distance effect). Bayes factors indicated data favored the null for each analysis, ranging from BF01 = 1.3 (retention) to 5.75 (performance). We found clear evidence of heterogeneity in each analysis, suggesting the impact of attentional focus depends on yet unknown contextual factors. Our results contradict the existing consensus that an external focus is always more effective than an internal focus. Instead, focus of attention appears to have a variety of effects that we cannot account for, and, on average, those effects are small to nil. These results parallel previous metascience suggesting publication bias has obfuscated the motor learning literature. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Team roles and challenges in experiential learning in sustainability accounting education: A four-stage learning cycle

ABSTRACT This study aims to provide insights into the process of team-based experiential learning in a sustainability accounting course. A case study of a postgraduate course on material flow cost accounting at a German university was applied. Data were collected from learning journals and in-depth interviews. The findings provide insights into the process of team-based experiential learning. The process is indicated by four team roles of students: interpersonal, information, analytic, and action roles, which move around the four-stage learning cycle. Six challenges in students' experiential learning in teams were found at different stages: the characteristics of team dynamics, the selection of relevant information from the case, confusion and frustration, connecting information in a meaningful manner, time constraints, and the increasingly complex level of the tasks. The study contributes to the accounting education literature and experiential learning literature by extending knowledge about team roles and learning challenges around the learning cycle. Practical implications for educators are discussed to facilitate team-based experiential learning toward sustainability education. These include the adoption of educator role profiles to support students' learning, the creation of a safe learning space for students, and the design of experiential learning activities on the foundation of four-stage learning processes.

Digital Employee Training With Digital Adoption Platforms Boost Learning and Knowledge Management of Corporate IT Systems

Due to increasing digitalization, employees need to use an increasing number of corporate IT systems. Digital Adoption Platforms (DAPs) recently address the resulting need for increased learning and skills to master such systems. Although of great practical importance, software adoption and learning literature has not yet studied the phenomenon of DAPs. Thus, after studying currently used DAPs (n=22 DAPs) we created our own DAP with interactive functionality based on this research. Subsequently we conducted an online experiment (n=152 employees) to assess if DAPs have positive training and learning effect on employees. More specifically, we compared the training and learning effects of using the DAP vs. using a classic user manual. The results of our ANOVA analyses show that employees using a DAP for learning will perceive a new corporate IT system as more useful than employees using a classic user manual for training. The use of DAPs significantly increases an employee's intention to use the new corporate IT system showing significant positive learning and training effects.

HIGHLY ADAPTIVE LASSO: MACHINE LEARNING THAT PROVIDES VALID NONPARAMETRIC INFERENCE IN REALISTIC MODELS.

Understanding treatment effects on health-related outcomes using real-world data requires defining a causal parameter and imposing relevant identification assumptions to translate it into a statistical estimand. Semiparametric methods, like the targeted maximum likelihood estimator (TMLE), have been developed to construct asymptotically linear estimators of these parameters. To further establish the asymptotic efficiency of these estimators, two conditions must be met: 1) the relevant components of the data likelihood must fall within a Donsker class, and 2) the estimates of nuisance parameters must converge to their true values at a rate faster than . The Highly Adaptive LASSO (HAL) satisfies these criteria by acting as an empirical risk minimizer within a class of càdlàg functions with a bounded sectional variation norm, which is known to be Donsker. HAL achieves the desired rate of convergence, thereby guaranteeing the estimators' asymptotic efficiency. The function class over which HAL minimizes its risk is flexible enough to capture realistic functions while maintaining the conditions for establishing efficiency. Additionally, HAL enables robust inference for non-pathwise differentiable parameters, such as the conditional average treatment effect (CATE) and causal dose-response curve, which are important in precision health. While these parameters are often considered in machine learning literature, these applications typically lack proper statistical inference. HAL addresses this gap by providing reliable statistical uncertainty quantification that is essential for informed decision-making in health research.

A Structural Topic and Sentiment-Discourse Model for Text Analysis

We consider the common setting where one observes a large number of opinionated text documents and related covariates, such as the text of online reviews along with the date of the review and the author demographic information. In this setting it can be of interest to understand how the covariates determine the text composition, as well as the prevalence, sentiment, and/or discourse of various discussion themes. Yet, most topic modeling methods in the machine learning literature are designed to summarize the text for the purpose of exploratory analysis and not to perform this type of formal statistical inference. Further, topic modeling methods generally do not try to estimate the sentiment or discourse of discussion along separate topics that can be critical in business applications (e.g., for summarizing service or product quality). We develop a topic model called the structural topic and sentiment-discourse (STS) model that introduces a new document-level latent variable that captures the sentiment and/or discourse (termed as “sentiment-discourse”) for each topic, which modulates the word frequency within a topic. These latent topic sentiment-discourse variables are controlled by document-level covariates to allow for experimental control and regression analysis. We also introduce new computational methods to resolve scalability issues that have forced previous models to restrict to a small number of categorical covariates. We benchmark the STS model on three real-world data sets from surveys, blogs, and Yelp restaurant reviews around the COVID-19 pandemic. Our model recovers meaningful results including rich insights about how COVID-19 affects online reviews, demonstrating that the STS model can be useful for regression analysis with text data in addition to topic modeling’s traditional use of descriptive analysis. This paper was accepted by Anindya Ghose, information systems. Supplemental Material: The online appendices and data files are available at https://doi.org/10.1287/mnsc.2022.00261 . An updated version of the R package implementing the STS model is available at https://CRAN.R-project.org/package=sts .

How the technologies behind self-driving cars, social networks, ChatGPT, and DALL-E2 are changing structural biology.

The performance of deep Neural Networks (NNs) in the text (ChatGPT) and image (DALL-E2) domains has attracted worldwide attention. Convolutional NNs (CNNs), Large Language Models (LLMs), Denoising Diffusion Probabilistic Models (DDPMs)/Noise Conditional Score Networks (NCSNs), and Graph NNs (GNNs) have impacted computer vision, language editing and translation, automated conversation, image generation, and social network management. Proteins can be viewed as texts written with the alphabet of amino acids, as images, or as graphs of interacting residues. Each of these perspectives suggests the use of tools from a different area of deep learning for protein structural biology. Here, I review how CNNs, LLMs, DDPMs/NCSNs, and GNNs have led to major advances in protein structure prediction, inverse folding, protein design, and small molecule design. This review is primarily intended as a deep learning primer for practicing experimental structural biologists. However, extensive references to the deep learning literature should also make it relevant to readers who have a background in machine learning, physics or statistics, and an interest in protein structural biology.

Doubly-robust and heteroscedasticity-aware sample trimming for causal inference

SUMMARY A popular method for variance reduction in causal inference is propensity-based trimming, the practice of removing units with extreme propensities from the sample. This practice has theoretical grounding when the data are homoscedastic and the propensity model is parametric (Yang & Ding, 2018; Crump et al., 2009), but in modern settings where heteroscedastic data are analyzed with non-parametric models, existing theory fails to support current practice. In this work, we address this challenge by developing new methods and theory for sample trimming. Our contributions are three-fold: first, we describe novel procedures for selecting which units to trim. Our procedures differ from previous works in that we trim not only units with small propensities, but also units with extreme conditional variances. Second, we give new theoretical guarantees for inference after trimming. In particular, we show how to perform inference on the trimmed sub-population without requiring that our regressions converge at parametric rates. Instead, we make only fourth-root rate assumptions like those in the double machine learning literature. This result applies to conventional propensity-based trimming as well and thus may be of independent interest. Finally, we propose a bootstrap-based method for constructing simultaneously valid confidence intervals for multiple trimmed sub-populations, which are valuable for navigating the trade-off between sample size and variance reduction inherent in trimming. We validate our methods in simulation, on the 2007-2008 National Health and Nutrition Examination Survey, and on a semi-synthetic Medicare dataset and find promising results in all settings.

Analysis of nailfold capillaroscopy images with artificial intelligence: Data from literature and performance of machine learning and deep learning from images acquired in the SCLEROCAP study

ObjectiveTo evaluate the performance of machine learning and then deep learning to detect a systemic scleroderma (SSc) landscape from the same set of nailfold capillaroscopy (NC) images from the French prospective multicenter observational study SCLEROCAP. MethodsNC images from the first 100 SCLEROCAP patients were analyzed to assess the performance of machine learning and then deep learning in identifying the SSc landscape, the NC images having previously been independently and consensually labeled by expert clinicians. Images were divided into a training set (70 %) and a validation set (30 %). After features extraction from the NC images, we tested six classifiers (random forests (RF), support vector machine (SVM), logistic regression (LR), light gradient boosting (LGB), extreme gradient boosting (XGB), K-nearest neighbors (KNN)) on the training set with five different combinations of the images. The performance of each classifier was evaluated by the F1 score. In the deep learning section, we tested three pre-trained models from the TIMM library (ResNet-18, DenseNet-121 and VGG-16) on raw NC images after applying image augmentation methods. ResultsWith machine learning, performance ranged from 0.60 to 0.73 for each variable, with Hu and Haralick moments being the most discriminating. Performance was highest with the RF, LGB and XGB models (F1 scores: 0.75–0.79). The highest score was obtained by combining all variables and using the LGB model (F1 score: 0.79 ± 0.05, p < 0.01). With deep learning, performance reached a minimum accuracy of 0.87. The best results were obtained with the DenseNet-121 model (accuracy 0.94 ± 0.02, F1 score 0.94 ± 0.02, AUC 0.95 ± 0.03) as compared to ResNet-18 (accuracy 0.87 ± 0.04, F1 score 0.85 ± 0.03, AUC 0.87 ± 0.04) and VGG-16 (accuracy 0.90 ± 0.03, F1 score 0.91 ± 0.02, AUC 0.91 ± 0.04). ConclusionBy using machine learning and then deep learning on the same set of labeled NC images from the SCLEROCAP study, the highest performances to detect SSc landscape were obtained with deep learning and in particular DenseNet-121. This pre-trained model could therefore be used to automatically interpret NC images in case of suspected SSc. This result nevertheless needs to be confirmed on a larger number of NC images.

Policy learning in crisis governance: translating narratives and reframing actions for co-evolving futures

ABSTRACT Most policy learning literature assumes identifiable lessons, comparable points of consideration, transferable solutions, and expectable performative changes. However, all these assumptions are challenged in the context of crises, where complexity makes learning across contexts difficult, if not impossible. This article aims to bring in policy learning from critical perspectives. It argues that, in the face of complexity, policy learning that can support crisis governance is instead an internal process involving the translation of lessons and the co-evolution of interrelated actors. In this lens, policy learning goes beyond the conventional understanding of lessons, drawing on actual practices within the learning loops of the particular policy cycle. Such loops and cycles, in this view, are not truly present. Grounded in Hyogo, Japan’s empowered participatory governance in responding to the crisis, policy learning is instead a complex everyday practice for translating narratives about past actions that, in turn, help frame future directions. With that framework, it can stimulate the co-evolution of state and societal actors without clear changing points, leading to more robust governing practices.

A Modal Type Theory of Expected Cost in Higher-Order Probabilistic Programs

The design of online learning algorithms typically aims to optimise the incurred loss or cost , e.g., the number of classification mistakes made by the algorithm. The goal of this paper is to build a type-theoretic framework to prove that a certain algorithm achieves its stated bound on the cost. Online learning algorithms often rely on randomness, their loss functions are often defined as expectations, precise bounds are often non-polynomial (e.g., logarithmic) and proofs of optimality often rely on potential-based arguments. Accordingly, we present pλ-amor, a type-theoretic graded modal framework for analysing (expected) costs of higher-order probabilistic programs with recursion. pλ-amor is an effect-based framework which uses graded modal types to represent potentials, cost and probability at the type level. It extends prior work (λ-amor) on cost analysis for deterministic programs. We prove pλ-amor sound relative to a Kripke step-indexed model which relates potentials with probabilistic coupling. We use pλ-amor to prove cost bounds of several examples from the online machine learning literature. Finally, we describe an extension of pλ-amor with a graded comonad and describe the relationship between the different modalities.

On the Possibilities of Poetry Turned into music. Experiences with Authors, Part of the Required and Elective Training Program

The proposed article examines the relationship between verse and music in the context of the popular Bulgarian song and the lyric poetry itself. Its aim is to offer a way of incorporating poetry in the classroom using one of the media for realisation of the language. The object of the analysis are poetic texts written by authors who are part of the required and elective school program. Thus, the proposed way of learning literature expands not only the spectrum of dimensions of seeing the arts but also the knowledge about the poets themselves.

Developing intercultural competence through literature in a multicultural framework

This article looks at how multilingualism is changing in today's world due to globalization, new technology, and more people moving across borders. It covers both the theory and practical aspects of multilingualism in communication, education, and society. Theoretically, it discusses the "multilingual turn" in education, which moves away from seeing monolingualism as the norm. It supports using flexible language practices like translanguaging, where different languages are mixed based on the situation. The practical side focuses on English as a global language and how multilingualism can be included in English as a Foreign Language (EFL) classes. It shows how using students’ different language backgrounds can improve learning. The article also highlights the importance of developing intercultural skills through literature, which helps students understand other cultures and reflect on their own. The methodology framework presented in the article is based on using literature in language learning to build cultural awareness and critical thinking.

A Trend Analysis of Project-based Learning in Chemistry Experiment: A Bibliometric Analysis

Project-based learning is a global trend, widely adopted across countries. Project-based learning has developed diversely by emphasizing interdisciplinary, cross-domain, and regional collaboration. To provide in-depth resources for researchers and educators on project-based learning in chemistry experiments, this study aims to make a bibliometric analysis on the topic of project-based learning globally using publication characteristics analysis. The database used in this study used Scopus and obtained 5900 articles from 2018-2023. Bibliometric analysis was analyzed using VOSviewer and Microsoft Excel. The results of the bibliometric analysis show that Boone, C.D. is the author with the highest number of citations on the research topic, Project-based learning in a chemistry experiment. The United States is the most productive country in researching this topic, with 1351 articles. The most influential journal with the highest number of articles is the Journal of Physics Conference Series, with 243 articles. The major contextualized learning literatures in this study tend to fall into broad groups such as Project-based Learning, Education, and STEM Education. These findings suggest the need for multidisciplinary and cross-disciplinary studies on project-based learning research on chemistry learning and advocate for the inclusion of project-based learning research in education from a broader geographical context.

Random Subspace Local Projections

Abstract We show how random subspace methods can be adapted to estimating local projections with many controls. Random subspace methods have their roots in the machine learning literature and are implemented by averaging over regressions estimated over different combinations of subsets of these controls. We document three key results: (i) Our approach can successfully recover the impulse response functions across Monte Carlo experiments representative of different macroeconomic settings and identification schemes. (ii) Our results suggest that random subspace methods are more accurate than other dimension reduction methods if the underlying large dataset has a factor structure similar to typical macroeconomic datasets such as FRED-MD. (iii) Our approach leads to differences in the estimated impulse response functions relative to benchmark methods when applied to two widely studied empirical applications.

Better-prepared to face COVID-19: Organizational learning from SMEs’ experience with banking crises

ABSTRACT This paper analyzes whether experiencing a banking crisis influenced firms’ performance during the COVID-19 pandemic. Drawing on the organizational learning and crisis management literature, we hypothesize that firms with banking crisis experience may learn from it and, thus, have a greater capacity to react and adapt to future external crises through adequate adaptative mechanisms. Using a survey sample of over 31,000 small- and medium-size enterprises operating in 39 countries, we obtain results consistent with our hypothesis that firms with banking crisis experience coped better during the pandemic than other firms. This finding is observed regardless of the severity of the banking crisis episode or how long ago the banking crisis occurred. We further demonstrate that banking crisis experience’s positive impact is explained partly by the increase in human resource slack before the pandemic as a coping strategy.

A systematic review of catching and comparing the most valuable gamification elements used in online learning

Learning through games has exploded worldwide. Many students enjoy learning through gaming programs, and its effects on curricula and instruction have recently increased. Saleem et al. (2022) recommended further research to investigate the most valuable elements of gamified online learning empirically. Thus, this study systematically explored the gamified elements employed in online learning literature, focusing on identifying the most practical and valuable elements and aiming to provide insights that contribute to the existing knowledge of gaming techniques in online education.

Scholarship of Teaching and Learning Research in Communication Sciences and Disorders Journals: An Exploratory Survey of the Recent Literature.

Inspired by a preliminary survey of the Scholarship of Teaching and Learning (SoTL) literature conducted by Friberg et al. (2014), the current study aimed to expand the original study's findings-that SoTL was rarely published in the communication sciences and disorders (CSD) literature from 2009 to 2013-to the subsequent 8-year period (2014-2021). The latter period was of particular relevance considering the dissolution of one American Speech-Language-Hearing Association (ASHA)-affiliated publication and the addition of another non-ASHA-affiliated publication during that time. Ten peer-reviewed CSD journals and the articles published in them were identified via a survey of the literature that used narrowly defined criteria developed in collaboration with a librarian. Five trained CSD graduate student raters compared a definition of SoTL to article abstracts to determine whether each should have been categorized as SoTL. Part 1 of the study investigated an 8-year time span, Part 2 investigated the remaining nine journals in the subsequent 5 years, and Part 3 investigated publication rates of SoTL in Teaching and Learning in Communication Sciences & Disorders (TLCSD) alone. Part 1 of the study revealed that publication rates of SoTL were extremely low in the 10 surveyed CSD journals in the 8-year time span investigated. A similar and diminishing trend was found in Part 2. Even smaller percentages of journals were dedicated to SoTL. Part 3 confirmed that relatively large amounts of CSD-specific SoTL have been published in TLCSD since its inception in 2017. As of 2021, SoTL articles continued to be uncommonly published in CSD journals. Until recently, scholarly teachers attempting to apply SoTL in the classroom had limited resources. Beginning in 2017, TLCSD has provided an outlet for SoTL, whereas other CSD journals seem to have published less of it. Reform of publication, peer review, tenure, and promotion policies and procedures is called for so that SoTL might be included as a valid scientific endeavor.