Parsimonious Representation Research Articles

Physics and chemistry from parsimonious representations: image analysis via invariant variational autoencoders

Electron, optical, and scanning probe microscopy methods are generating ever increasing volume of image data containing information on atomic and mesoscale structures and functionalities. This necessitates the development of the machine learning methods for discovery of physical and chemical phenomena from the data, such as manifestations of symmetry breaking phenomena in electron and scanning tunneling microscopy images, or variability of the nanoparticles. Variational autoencoders (VAEs) are emerging as a powerful paradigm for the unsupervised data analysis, allowing to disentangle the factors of variability and discover optimal parsimonious representation. Here, we summarize recent developments in VAEs, covering the basic principles and intuition behind the VAEs. The invariant VAEs are introduced as an approach to accommodate scale and translation invariances present in imaging data and separate known factors of variations from the ones to be discovered. We further describe the opportunities enabled by the control over VAE architecture, including conditional, semi-supervised, and joint VAEs. Several case studies of VAE applications for toy models and experimental datasets in Scanning Transmission Electron Microscopy are discussed, emphasizing the deep connection between VAE and basic physical principles. Python codes and datasets discussed in this article are available at https://github.com/saimani5/VAE-tutorials and can be used by researchers as an application guide when applying these to their own datasets.

Multiple pandemic waves vs multi-period/multi-phasic epidemics: Global shape of the COVID-19 pandemic

The overall course of the COVID-19 pandemic in Western countries has been characterized by complex sequences of phases. In the period before the arrival of vaccines, these phases were mainly due to the alternation between the strengthening/lifting of social distancing measures, with the aim to balance the protection of health and that of the society as a whole. After the arrival of vaccines, this multi-phasic character was further emphasized by the complicated deployment of vaccination campaigns and the onset of virus’ variants. To cope with this multi-phasic character, we propose a theoretical approach to the modeling of overall pandemic courses, that we term multi-period/multi-phasic, based on a specific definition of phase. This allows a unified and parsimonious representation of complex epidemic courses even when vaccination and virus’ variants are considered, through sequences of weak ergodic renewal equations that become fully ergodic when appropriate conditions are met. Specific hypotheses on epidemiological and intervention parameters allow reduction to simple models. The framework suggest a simple, theory driven, approach to data explanation that allows an accurate reproduction of the overall course of the COVID-19 epidemic in Italy since its beginning (February 2020) up to omicron onset, confirming the validity of the concept.

BAYESIAN NESTED LATENT CLASS MODELS FOR CAUSE-OF-DEATH ASSIGNMENT USING VERBAL AUTOPSIES ACROSS MULTIPLE DOMAINS.

Understanding cause-specific mortality rates is crucial for monitoring population health and designing public health interventions. Worldwide, two-thirds of deaths do not have a cause assigned. Verbal autopsy (VA) is a well-established tool to collect information describing deaths outside of hospitals by conducting surveys to caregivers of a deceased person. It is routinely implemented in many low- and middle-income countries. Statistical algorithms to assign cause of death using VAs are typically vulnerable to the distribution shift between the data used to train the model and the target population. This presents a major challenge for analyzing VAs, as labeled data are usually unavailable in the target population. This article proposes a latent class model framework for VA data (LCVA) that jointly models VAs collected over multiple heterogeneous domains, assigns causes of death for out-of-domain observations and estimates cause-specific mortality fractions for a new domain. We introduce a parsimonious representation of the joint distribution of the collected symptoms using nested latent class models and develop a computationally efficient algorithm for posterior inference. We demonstrate that LCVA outperforms existing methods in predictive performance and scalability. Supplementary Material and reproducible analysis codes are available online. The R package LCVA implementing the method is available on GitHub (https://github.com/richardli/LCVA).

A Multicountry Model of the Term Structures of Interest Rates with a GVAR

Abstract Extant multicountry affine term structure models (ATSMs) handle global financial interdependence at the cost of increasing model dimensionality. To address this challenge, we propose a novel no-arbitrage ATSM with risk factor dynamics following a global vector-autoregressive (GVAR). Compared to referenced benchmarks, the GVAR − ATSM offers a more parsimonious representation, enables a faster estimation process, produces more precise model estimates, yields more plausible term premia dynamics, and improves bond yield out-of-sample forecasting. Furthermore, our empirical findings reveal the significant impact of China’s economic stances on Latin American yield curve dynamics, underscoring its importance as a global economic player.

Underground Diagnosis Based on GPR and Learning in the Model Space.

Ground Penetrating Radar (GPR) has been widely used in pipeline detection and underground diagnosis. In practical applications, the characteristics of the GPR data of the detected area and the likely underground anomalous structures could be rarely acknowledged before fully analyzing the obtained GPR data, causing challenges to identify the underground structures or anomalies automatically. In this article, a GPR B-scan image diagnosis method based on learning in the model space is proposed. The idea of learning in the model space is to use models fitted on parts of data as more stable and parsimonious representations of the data. For the GPR image, 2-Direction Echo State Network (2D-ESN) is proposed to fit the image segments through the next item prediction. By building the connections between the points on the image in both the horizontal and vertical directions, the 2D-ESN regards the GPR image segment as a whole and could effectively capture the dynamic characteristics of the GPR image. And then, semi-supervised and supervised learning methods could be further implemented on the 2D-ESN models for underground diagnosis. Experiments on real-world datasets are conducted, and the results demonstrate the effectiveness of the proposed model.

Psychometric properties of the Italian version of the Occupational Fatigue Exhaustion/Recovery (OFER) scale among healthcare professionals

ABSTRACT Objective The purpose of this study was to analyze the psychometric properties of the Italian version of the Occupational Fatigue Exhaustion/Recovery (OFER). Methods A cross-sectional study among 434 healthcare workers was undertaken. The factorial validity of the OFER, was compared using standard confirmatory factor analysis (CFA) with exploratory structural equation modeling (ESEM). Internal consistency, convergent validity, and factorial invariance (across gender) were assessed. Results A three-factor first-order ESEM solution provided the best and most parsimonious representation of the data: χ2(63) = 212.81, p < .001, RMSEA = 0.074 (90% CI 0.063–0.085), CFI = 0.949, TLI = 0.915. The three-factors were chronic fatigue, acute fatigue, and intershift recovery. Chronic and acute factors showed excellent internal consistency (ω > 0.80), however intershift recovery showed less internal consistency (ω < 0.70). OFER scores were found to be positively correlated with FAS scores, supporting its convergent validity. Conclusions The OFER was factorially invariant across gender. Overall, our findings provided evidence for the satisfactory psychometric properties of the OFER among Italian workers, confirming its value for use in both practical occupational and academic research settings.

Leveraging language representation for materials exploration and discovery

Data-driven approaches to materials exploration and discovery are building momentum due to emerging advances in machine learning. However, parsimonious representations of crystals for navigating the vast materials search space remain limited. To address this limitation, we introduce a materials discovery framework that utilizes natural language embeddings from language models as representations of compositional and structural features. The contextual knowledge encoded in these language representations conveys information about material properties and structures, enabling both similarity analysis to recall relevant candidates based on a query material and multi-task learning to share information across related properties. Applying this framework to thermoelectrics, we demonstrate diversified recommendations of prototype crystal structures and identify under-studied material spaces. Validation through first-principles calculations and experiments confirms the potential of the recommended materials as high-performance thermoelectrics. Language-based frameworks offer versatile and adaptable embedding structures for effective materials exploration and discovery, applicable across diverse material systems.

Toward the development of an ecosystem model ensemble to support adaptive management in Lake Ontario

Notwithstanding the continuing advancement of our understanding of the broader ecosystem functioning in Lake Ontario, emerging evidence suggests that there are fundamental knowledge gaps to accurately describe the relationship between exogenous phosphorus (P) loading and in-lake total phosphorus (TP) concentrations. The whole-lake load appears to frequently exceed the Great Lakes Water Quality Agreement (GLWQA) interim target of 7000 metric tonnes·year−1, although the offshore-water P concentrations are consistently below the GLWQA interim substance concentration objective of 10 µg TP·L–1. Contrasting ecological conditions prevail in different locations of Lake Ontario, owing to the significant urban footprint along the shoreline but also the capacity of dreissenid mussels to sequester P in the littoral zone. Specifically, low ambient P levels threaten fisheries productivity in offshore waters, while nuisance benthic algae (predominantly Cladophora) and toxin-producing cyanobacteria blooms affect the water quality in the nearshore zone. The present study offers a technical analysis of the recent and ongoing modelling work that has been conducted in Lake Ontario and can be potentially used to address the multitude of ecosystem management challenges. Our aim is to provide an overview of all the major models developed by identifying their fundamental assumptions, structural attributes, and general consistency against empirical knowledge derived from the system. The existing modelling work has opted for parsimonious representations of the lower food web coupled with granular grid configurations to effectively link hydrodynamic processes and mass transport between nearshore and offshore waters. The establishment of comprehensive ecophysiological modules that will recreate the mechanisms underlying the interplay among bioavailable phosphorus, planktonic dynamics, dreissenid mussels, and Cladophora is a critical undertaking to reproduce the water quality conditions in the nearshore zone of Lake Ontario. Striving to integrate the lower food web with fisheries and ecosystem-service modelling, we also offer a technical analysis of knowledge gaps and monitoring-assessment objectives that should be addressed to ensure that ecosystem processes of management interest are adequately measured and the local modelling enterprise is focused on suitable performance indicators.

Quantics Tensor Cross Interpolation for High-Resolution Parsimonious Representations of Multivariate Functions.

Multivariate functions of continuous variables arise in countless branches of science. Numerical computations with such functions typically involve a compromise between two contrary desiderata: accurate resolution of the functional dependence, versus parsimonious memory usage. Recently, two promising strategies have emerged for satisfying both requirements: (i)The quantics representation, which expresses functions as multi-index tensors, with each index representing one bit of a binary encoding of one of the variables; and (ii)tensor cross interpolation (TCI), which, if applicable, yields parsimonious interpolations for multi-index tensors. Here, we present a strategy, quantics TCI, which combines the advantages of both schemes. We illustrate its potential with an application from condensed matter physics: the computation of Brillouin zone integrals.

Detection of common subtrees with identical label distribution

Frequent pattern mining is a relevant method to analyse structured data, like sequences, trees or graphs. It consists in identifying characteristic substructures of a dataset. This paper deals with a new type of patterns for tree data: common subtrees with identical label distribution. Their detection is far from obvious since the underlying isomorphism problem is graph isomorphism complete. An elaborated search algorithm is developed and analysed from both theoretical and numerical perspectives. Based on this, the enumeration of patterns is performed through a new lossless compression scheme for trees, called DAG-RW, whose complexity is investigated as well. The method shows very good properties, both in terms of computation times and analysis of real datasets from the literature. Compared to other substructures like topological subtrees and labelled subtrees for which the isomorphism problem is linear, the patterns found provide a more parsimonious representation of the data.

Efficient in-situ image and video compression through probabilistic image representation

Fast and effective image compression for multi-dimensional images has become increasingly important for efficient storage and transfer of massive amounts of high-resolution images and videos. In this paper, we present an efficient in-situ method for multi-dimensional image and video compression called Compression via Adaptive Recursive Partitioning (CARP). CARP uses an optimal permutation of the image pixels inferred from a Bayesian probabilistic model on recursive partitions of the image to reduce its effective dimensionality, achieving a parsimonious representation that preserves information. Furthermore, it adopts a multi-layer Bayesian hierarchical model to achieve in-situ compression along with self-tuning and regularization, with just one single parameter to be specified by the user to achieve the desired compression rate. The properties of our proposed method include high reconstruction quality at a wide range of compression rates while preserving key local details, applicability to a variety of different image/video types and of different dimensions, computational scalability, progressive transmission and ease of tuning. Extensive numerical experiments using a variety of datasets including 2D still images, real-life YouTube videos, and surveillance videos show that CARP compares favorably to—and often uniformly outperforms—a wide range of popular image/video compression approaches, including JPEG, JPEG2000, AVI, BPG, MPEG4, HEVC, AV1, and three neural network-based methods.

PEPPR: A post-encoding pre-production reinstatement model of dual-list free recall.

Recent events are easy to recall, but they also interfere with the recall of more distant, non-recent events. In many computational models, non-recent memories are recalled by using the context associated with those events as a cue. Some models, however, do little to explain how people initially activate non-recent contexts in the service of accurate recall. We addressed this limitation by evaluating two candidate mechanisms within the Context-Maintenance and Retrieval model. The first is a Backward-Walk mechanism that iteratively applies a generate/recognize process to covertly retrieve progressively less recent items. The second is a Post-Encoding Pre-Production Reinstatement (PEPPR) mechanism that formally implements a metacognitive control process that reinstates non-recent contexts prior to retrieval. Models including these mechanisms make divergent predictions about the dynamics of response production and monitoring when recalling non-recent items. Before producing non-recent items, Backward-Walk cues covert retrievals of several recent items, whereas PEPPR cues few, if any, covert retrievals of that sort. We tested these predictions using archival data from a dual-list externalized free recall paradigm that required subjects to report all items that came to mind while recalling from the non-recent list. Simulations showed that only the model including PEPPR accurately predicted covert recall patterns. That same model fit the behavioral data well. These findings suggest that self-initiated context reinstatement plays an important role in recall of non-recent memories and provides a formal model that uses a parsimonious non-hierarchical context representation of how such reinstatement might occur.

Cross-cultural adaptation and psychometric properties of the Italian version of the Orthorexia Nervosa Inventory (ONI)

BackgroundOrthorexia Nervosa (ON) is an emerging clinical condition characterized by a pathological fixation with healthy eating. Recently, the Orthorexia Nervosa Inventory (ONI) has emerged as a promising tool for assessing orthorexic tendencies and behaviours, aiming to overcome the well-established limitations of existing measures for ON. The present study aimed to examine the psychometric properties of the Italian version of the ONI.MethodsA total of 879 participants (Mage = 33.22 years, SD = 9.19; 56.9% females) completed the ONI along with the Düsseldorf Orthorexia Scale (DOS), measures of disordered eating, obsessive-compulsive symptoms (OCD), and psychosocial impairment. To establish the factorial validity of the ONI, a competing measurement modeling approach was employed by comparing standard confirmatory factor analytic (CFA) with exploratory structural equation modeling (ESEM) solutions. Model-based omega coefficients were computed to examine the internal consistency of the scale. Factorial invariance tests across gender were conducted within a multi-group framework.ResultsA three-factor first-order ESEM solution provided the best and most parsimonious representation of the data: χ2(207) = 558.641, p < .001, RMSEA = 0.044 (90% CI 0.040–0.048), CFI = 0.976, TLI = 0.968, SRMR = 0.036. The three latent dimensions, labelled behaviors, emotions, and impairments, showed excellent internal consistency (ω > 0.88). Furthermore, ONI scores were found to be positively correlated with DOS scores, disordered eating, OCD symptoms, and psychosocial impairment, supporting its convergent and criterion validity. Eventually, the ONI was factorially invariant across gender.ConclusionsOverall, the present study provides evidence for the satisfactory psychometric properties of the ONI in the Italian context, endorsing its use in both clinical and research settings.

Gaussian hierarchical latent Dirichlet allocation: Bringing polysemy back.

Topic models are widely used to discover the latent representation of a set of documents. The two canonical models are latent Dirichlet allocation, and Gaussian latent Dirichlet allocation, where the former uses multinomial distributions over words, and the latter uses multivariate Gaussian distributions over pre-trained word embedding vectors as the latent topic representations, respectively. Compared with latent Dirichlet allocation, Gaussian latent Dirichlet allocation is limited in the sense that it does not capture the polysemy of a word such as "bank." In this paper, we show that Gaussian latent Dirichlet allocation could recover the ability to capture polysemy by introducing a hierarchical structure in the set of topics that the model can use to represent a given document. Our Gaussian hierarchical latent Dirichlet allocation significantly improves polysemy detection compared with Gaussian-based models and provides more parsimonious topic representations compared with hierarchical latent Dirichlet allocation. Our extensive quantitative experiments show that our model also achieves better topic coherence and held-out document predictive accuracy over a wide range of corpus and word embedding vectors which significantly improves the capture of polysemy compared with GLDA and CGTM. Our model learns the underlying topic distribution and hierarchical structure among topics simultaneously, which can be further used to understand the correlation among topics. Moreover, the added flexibility of our model does not necessarily increase the time complexity compared with GLDA and CGTM, which makes our model a good competitor to GLDA.

Combining discrete choice models and neural networks through embeddings: Formulation, interpretability and performance

This paper proposes a novel approach that combines theory- and data-driven choice models using Artificial Neural Networks (ANNs). In particular, we use continuous vector representations, called embeddings, for encoding categorical or discrete explanatory variables with a special focus on interpretability and model transparency. Although embedding representations within the logit framework have been conceptualized by Pereira (2019), their dimensions do not have an absolute definitive meaning, hence offering limited behavioral insights in this earlier work. The novelty of our work lies in enforcing interpretability to the embedding vectors by formally associating each of their dimensions to a choice alternative. Thus, our approach brings benefits much beyond a simple parsimonious representation improvement over dummy encoding, as it provides behaviorally meaningful outputs that can be used in travel demand analysis and policy decisions. Additionally, in contrast to previously suggested ANN-based Discrete Choice Models (DCMs) that either sacrifice interpretability for performance or are only partially interpretable, our main model (E-MNL) preserves interpretability of the utility coefficients for all the input variables despite being based on ANN principles. The proposed models are tested on two real world datasets and evaluated against benchmark and baseline models that use dummy-encoding. The results of the experiments indicate that our models deliver state-of-the-art predictive performance, outperforming existing ANN-based models while drastically reducing the number of required network parameters.

The Timing of Global Floods and Its Association With Climate and Topography

AbstractUntil recently, the development of a global geography of floods was challenged by the fragmentation and heterogeneity of in situ data and the high costs of processing large amounts of remote sensing data. Such geography would facilitate the exploration of large‐scale drivers of flood extent and timing including wide latitudinal, climate, and topographic effects. Here we used a monthly data set spanning 30 years (Global Surface Water) to develop a worldwide geographical characterization of slow floods (1‐degree grid), weighting the relative contribution of seasonal, interannual, and long‐term fluctuations on overall variability, and quantified precipitation‐flooding delays where seasonality dominated. We explored the dominance of different flooding timings across five main Köppen‐Geiger climates and seven topography classes derived from modeled water table depths (i.e., hydro‐topography) to contribute top‐down insight about the most salient, cross‐regional flooding patterns and their likely large‐scale drivers. Our results showed that, globally, the mean extent of floods averaged 0.48% of the global land area, predominantly associated with hydro‐topography (&gt;2× more extensive in flatter landscapes). Climate drove flood timings, with predictable, seasonally dominated fluctuations in cold regions, interannual, and mixed patterns in temperate climates, and more irregular (higher variability) and unpredictable (less seasonal) patterns in arid regions. Net gains of flooded area dominated temporal variability in 9% of the cells including boreal clusters likely affected by warming trends. We propose that this new geographical perspective of floods can aid different avenues of hydrological research in the upscaling and extrapolation of field studies and the parsimonious representation of floods in hydroclimatic models.

GRIDS-Net: Inverse shape design and identification of scatterers via geometric regularization and physics-embedded deep learning

This study presents a deep learning based methodology for both remote sensing and design of acoustic scatterers. The ability to determine the shape of a scatterer, either in the context of material design or sensing, plays a critical role in many practical engineering problems. This class of inverse problems is extremely challenging due to their high-dimensional, nonlinear, and ill-posed nature. To overcome these technical hurdles, we introduce a geometric regularization approach for deep neural networks (DNN) based on non-uniform rational B-splines (NURBS) and capable of predicting complex 2D scatterer geometries in a parsimonious dimensional representation. Then, this geometric regularization is combined with physics-embedded learning and integrated within a robust convolutional autoencoder (CAE) architecture to accurately predict the shape of 2D scatterers in the context of identification and inverse design problems. An extensive numerical study is presented in order to showcase the remarkable ability of this approach to handle complex scatterer geometries while generating physically-consistent acoustic fields. The study also assesses and contrasts the role played by the (weakly) embedded physics in the convergence of the DNN predictions to a physically consistent inverse design.

Enhancing Cooperation of Vehicle Merging Control in Heavy Traffic Using Communication-Based Soft Actor-Critic Algorithm

A promising way to improve efficiency in highway on-ramp regions is to control connected and automated vehicles (CAVs) to pass the merging section sequentially. The primary objective of this paper is to incorporate the reinforcement learning (RL) technique into vehicle merging control to achieve global cooperation. A communication protocol among RL agents is integrated with the Soft Actor-Critic (CSAC) algorithm. The parallel SAC agents with a parameter-sharing structure cooperate to optimize the common reward function. It enables CAVs to know the actions of each other so that they proactively negotiate to adjust speeds. We designed a parsimonious state representation containing crucial merging information to speed up the RL training. The effects of the merging control strategy were investigated with the simulation model in scenarios with different CAV penetration rates and traffic flow compositions. Results showed that the CSAC-based merging strategy generated collision-free merging trajectories with a short travel time while guaranteeing traffic safety in various traffic conditions. Four baselines (including two rule-based and two RL-based merging strategies) were applied in the same scenarios for comparison. It was found that the proposed merging strategy took the safest merging behaviors (zero traffic conflicts) and reduced the most significant travel time (56.9%).

Humans parsimoniously represent auditory sequences by pruning and completing the underlying network structure.

Successive auditory inputs are rarely independent, their relationships ranging from local transitions between elements to hierarchical and nested representations. In many situations, humans retrieve these dependencies even from limited datasets. However, this learning at multiple scale levels is poorly understood. Here, we used the formalism proposed by network science to study the representation of local and higher-order structures and their interaction in auditory sequences. We show that human adults exhibited biases in their perception of local transitions between elements, which made them sensitive to high-order network structures such as communities. This behavior is consistent with the creation of a parsimonious simplified model from the evidence they receive, achieved by pruning and completing relationships between network elements. This observation suggests that the brain does not rely on exact memories but on a parsimonious representation of the world. Moreover, this bias can be analytically modeled by a memory/efficiency trade-off. This model correctly accounts for previous findings, including local transition probabilities as well as high-order network structures, unifying sequence learning across scales. We finally propose putative brain implementations of such bias.

Can we pull resilience from the rubble? experiencing earthquakes in Mexico City

This paper serves as an introduction to focus on the meaning of resilience as a notion that hypothetically enables cities and their citizens to remain unaffected by shocks and stresses of diverse nature, but in practice, is a parsimonious representation of a highly complex human endogenous condition that is often not adequately understood and addressed due to weak or weathered disaster risk governance. Through an empirical analysis, the relevance of the occurrence of disasters such as the one produced by the Mexico City earthquake of September 19, 2017, and its relationship to the meaning of urban resilience is addressed. Since the 100 Resilient Cities program launch in 2013, several cities, including Mexico City, have adopted diverse strategies to build and become more resilient to physical, social, and economic shocks and stresses. Nonetheless, the degree of success of such strategies expressed in everyday life within the cities, regardless of what the official documents report, reveals highly heterogeneous results. In the case of Mexico City, it is evident that the appropriate strategy for attaining the desired outcome of becoming a Resilient City has not been successful. Despite setting the goal of advancing disaster risk reduction through urban and regional planning, the impact of the earthquake of September 19, 2017, proved otherwise.