Reproducibility in Matrix and Tensor Decompositions: Focus on model match, interpretability, and uniqueness

Kantar, Furkan; Akhonda, M. A. B. S.; Strother, Stephen C.; Calhoun, Vince D.; Acar, Evrim

doi:10.1109/msp.2022.3163870

Cited by 27 publications

(30 citation statements)

References 73 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As such, it provides an effective summary of interactions between functional networks. In [2 ▪▪ ], the desirable match of the ICA model for fMRI analysis is demonstrated with an example using FNC maps, which we reproduce in Fig. 1.…”

Section: Two Case Studies For Reproducibility: Independent Component ...mentioning

confidence: 79%

“…Since in most cases, closed form solutions do not exist, iterative techniques are employed, and most commonly, using random initializations, see for example, [6 ▪ ,7,8]. Even when all algorithmic quantities are fixed, and the only variability is due to random initializations, the resulting decompositions can be quite different as demonstrated in [2 ▪▪ ]. Hence, for reproducibility, we should select an appropriate metric such as correlation to measure the consistency (repeatability, stability) of the final results.…”

Section: Definitions: Reproducibility and Replicabilitymentioning

confidence: 99%

“…1. GIFT is used to perform ICA analysis using the entropy bound minimization (EBM) algorithm with 176 subjects (88 healthy controls (HCs) and 88 patients with schizophrenia) from https://coins.trendscenter.org/, and 28 components were selected as functionally relevant (see [2 ▪▪ ] for experimental details). The FNC maps corresponding to the run with the lowest Cross-ISI (best run) and the one to lowest Cross-ISI among 100 runs with random initializations are shown in the figure.…”

Section: Two Case Studies For Reproducibility: Independent Component ...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…A recent overview, [2 ▪▪ ], starts with these definitions and underlines the importance of reproducibility as the starting point for all studies, including those on replicability. The focus in [2 ▪▪ ] is on unsupervised machine learning methods based on matrix and tensor decompositions (MTDs), which have found fruitful application in NI data analysis. In this review, we revisit the conclusions of [2 ▪▪ ] with an emphasis on applications in neuroscience, especially of independent component analysis (ICA) and review (the limited) work that addresses reproducibility in other machine learning solutions such as iterative optimization approaches including computational morphometry implemented in FreeSurfer…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Reproducibility and replicability in neuroimaging data analysis

Calhoun

2022

Current Opinion in Neurology

Self Cite

View full text Add to dashboard Cite

Purpose of reviewMachine learning solutions are being increasingly used in the analysis of neuroimaging (NI) data, and as a result, there is an increase in the emphasis of the reproducibility and replicability of these data-driven solutions. Although this is a very positive trend, related terminology is often not properly defined, and more importantly, (computational) reproducibility that refers to obtaining consistent results using the same data and the same code is often disregarded.Recent findingsWe review the findings of a recent paper on the topic along with other relevant literature, and present two examples that demonstrate the importance of accounting for reproducibility in widely used software for NI data.SummaryWe note that reproducibility should be a first step in all NI data analyses including those focusing on replicability, and introduce available solutions for assessing reproducibility. We add the cautionary remark that when not taken into account, lack of reproducibility can significantly bias all subsequent analysis stages.

show abstract

Section: Two Case Studies For Reproducibility: Independent Component ...mentioning

confidence: 79%

Section: Definitions: Reproducibility and Replicabilitymentioning

confidence: 99%

Section: Two Case Studies For Reproducibility: Independent Component ...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Reproducibility and replicability in neuroimaging data analysis

Calhoun

2022

Current Opinion in Neurology

Self Cite

View full text Add to dashboard Cite

show abstract

UnsupervisedEHR‐based phenotyping via matrix and tensor decompositions

Becker

Smilde

Acar

2023

WIREs Data Min & Knowl

Self Cite

View full text Add to dashboard Cite

Computational phenotyping allows for unsupervised discovery of subgroups of patients as well as corresponding co‐occurring medical conditions from electronic health records (EHR). Typically, EHR data contains demographic information, diagnoses and laboratory results. Discovering (novel) phenotypes has the potential to be of prognostic and therapeutic value. Providing medical practitioners with transparent and interpretable results is an important requirement and an essential part for advancing precision medicine. Low‐rank data approximation methods such as matrix (e.g., nonnegative matrix factorization) and tensor decompositions (e.g., CANDECOMP/PARAFAC) have demonstrated that they can provide such transparent and interpretable insights. Recent developments have adapted low‐rank data approximation methods by incorporating different constraints and regularizations that facilitate interpretability further. In addition, they offer solutions for common challenges within EHR data such as high dimensionality, data sparsity and incompleteness. Especially extracting temporal phenotypes from longitudinal EHR has received much attention in recent years. In this paper, we provide a comprehensive review of low‐rank approximation‐based approaches for computational phenotyping. The existing literature is categorized into temporal versus static phenotyping approaches based on matrix versus tensor decompositions. Furthermore, we outline different approaches for the validation of phenotypes, that is, the assessment of clinical significance. This article is categorized under: Algorithmic Development > Structure Discovery Fundamental Concepts of Data and Knowledge > Explainable AI Technologies > Machine Learning

show abstract

Characterizing human postprandial metabolic response using multiway data analysis

Yan,

Li,

Horner

et al. 2024

Metabolomics

View full text Add to dashboard Cite

Introduction Analysis of time-resolved postprandial metabolomics data can improve our understanding of the human metabolism by revealing similarities and differences in postprandial responses of individuals. Traditional data analysis methods often rely on data summaries or univariate approaches focusing on one metabolite at a time. Objectives Our goal is to provide a comprehensive picture in terms of the changes in the human metabolism in response to a meal challenge test, by revealing static and dynamic markers of phenotypes, i.e., subject stratifications, related clusters of metabolites, and their temporal profiles. Methods We analyze Nuclear Magnetic Resonance (NMR) spectroscopy measurements of plasma samples collected during a meal challenge test from 299 individuals from the COPSAC2000 cohort using a Nightingale NMR panel at the fasting and postprandial states (15, 30, 60, 90, 120, 150, 240 min). We investigate the postprandial dynamics of the metabolism as reflected in the dynamic behaviour of the measured metabolites. The data is arranged as a three-way array: subjects by metabolites by time. We analyze the fasting state data to reveal static patterns of subject group differences using principal component analysis (PCA), and fasting state-corrected postprandial data using the CANDECOMP/PARAFAC (CP) tensor factorization to reveal dynamic markers of group differences. Results Our analysis reveals dynamic markers consisting of certain metabolite groups and their temporal profiles showing differences among males according to their body mass index (BMI) in response to the meal challenge. We also show that certain lipoproteins relate to the group difference differently in the fasting vs. dynamic state. Furthermore, while similar dynamic patterns are observed in males and females, the BMI-related group difference is observed only in males in the dynamic state. Conclusion The CP model is an effective approach to analyze time-resolved postprandial metabolomics data, and provides a compact but a comprehensive summary of the postprandial data revealing replicable and interpretable dynamic markers crucial to advance our understanding of changes in the metabolism in response to a meal challenge.

show abstract

Reproducibility in Matrix and Tensor Decompositions: Focus on model match, interpretability, and uniqueness

Cited by 27 publications

References 73 publications

Reproducibility and replicability in neuroimaging data analysis

Reproducibility and replicability in neuroimaging data analysis

UnsupervisedEHR‐based phenotyping via matrix and tensor decompositions

Characterizing human postprandial metabolic response using multiway data analysis

Contact Info

Product

Resources

About