Browse
Submit Data
Databases
API
Help

Dataset Information

0 Views

0 Connections

0 Citations

0 Reanalyses

0 Downloads

Omics score: 0

Domain Generalization via Multidomain Discriminant Analysis.

ABSTRACT: Domain generalization (DG) aims to incorporate knowledge from multiple source domains into a single model that could generalize well on unseen target domains. This problem is ubiquitous in practice since the distributions of the target data may rarely be identical to those of the source data. In this paper, we propose Multidomain Discriminant Analysis (MDA) to address DG of classification tasks in general situations. MDA learns a domain-invariant feature transformation that aims to achieve appealing properties, including a minimal divergence among domains within each class, a maximal separability among classes, and overall maximal compactness of all classes. Furthermore, we provide the bounds on excess risk and generalization error by learning theory analysis. Comprehensive experiments on synthetic and real benchmark datasets demonstrate the effectiveness of MDA.

SUBMITTER: Hu S

PROVIDER: S-EPMC6759816 | biostudies-literature | 2019 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Json Xml

Publications

Domain Generalization via Multidomain Discriminant Analysis.

Hu Shoubo S Zhang Kun K Chen Zhitang Z Chan Laiwan L

Uncertainty in artificial intelligence : proceedings of the ... conference. Conference on Uncertainty in Artificial Intelligence 20190701

Domain generalization (DG) aims to incorporate knowledge from multiple source domains into a single model that could generalize well on unseen target domains. This problem is ubiquitous in practice since the distributions of the target data may rarely be identical to those of the source data. In this paper, we propose Multidomain Discriminant Analysis (MDA) to address DG of classification tasks in general situations. MDA learns a domain-invariant feature transformation that aims to achieve appea ...[more]

PMID: 31555066

Similar Datasets

Fairness and Accuracy Under Domain Generalization.

Project description:As machine learning (ML) algorithms are increasingly used in high-stakes applications, concerns have arisen that they may be biased against certain social groups. Although many approaches have been proposed to make ML models fair, they typically rely on the assumption that data distributions in training and deployment are identical. Unfortunately, this is commonly violated in practice and a model that is fair during training may lead to an unexpected outcome during its deployment. Although the problem of designing robust ML models under dataset shifts has been widely studied, most existing works focus only on the transfer of accuracy. In this paper, we study the transfer of both fairness and accuracy under domain generalization where the data at test time may be sampled from never-before-seen domains. We first develop theoretical bounds on the unfairness and expected loss at deployment, and then derive sufficient conditions under which fairness and accuracy can be perfectly transferred via invariant representation learning. Guided by this, we design a learning algorithm such that fair ML models learned with training data still have high fairness and accuracy when deployment environments change. Experiments on real-world data validate the proposed algorithm. Model implementation is available at https://github.com/pth1993/FATDM.

| S-EPMC10246117 | biostudies-literature

Tandem domain swapping: determinants of multidomain protein misfolding.

Project description:Tandem homologous domains in proteins are susceptible to misfolding through the formation of domain swaps, non-native conformations involving the exchange of equivalent structural elements between adjacent domains. Cutting-edge biophysical experiments have recently allowed the observation of tandem domain swapping events at the single molecule level. In addition, computer simulations have shed light into the molecular mechanisms of domain swap formation and serve as the basis for methods to systematically predict them. At present, the number of studies on tandem domain swaps is still small and limited to a few domain folds, but they offer important insights into the folding and evolution of multidomain proteins with applications in the field of protein design.

| S-EPMC6863430 | biostudies-literature

Covariance-enhanced discriminant analysis.

Project description:Linear discriminant analysis has been widely used to characterize or separate multiple classes via linear combinations of features. However, the high dimensionality of features from modern biological experiments defies traditional discriminant analysis techniques. Possible interfeature correlations present additional challenges and are often underused in modelling. In this paper, by incorporating possible interfeature correlations, we propose a covariance-enhanced discriminant analysis method that simultaneously and consistently selects informative features and identifies the corresponding discriminable classes. Under mild regularity conditions, we show that the method can achieve consistent parameter estimation and model selection, and can attain an asymptotically optimal misclassification rate. Extensive simulations have verified the utility of the method, which we apply to a renal transplantation trial.

| S-EPMC4912060 | biostudies-literature

Kernel optimization in discriminant analysis.

Project description:Kernel mapping is one of the most used approaches to intrinsically derive nonlinear classifiers. The idea is to use a kernel function which maps the original nonlinearly separable problem to a space of intrinsically larger dimensionality where the classes are linearly separable. A major problem in the design of kernel methods is to find the kernel parameters that make the problem linear in the mapped representation. This paper derives the first criterion that specifically aims to find a kernel representation where the Bayes classifier becomes linear. We illustrate how this result can be successfully applied in several kernel discriminant analysis algorithms. Experimental results, using a large number of databases and classifiers, demonstrate the utility of the proposed approach. The paper also shows (theoretically and experimentally) that a kernel version of Subclass Discriminant Analysis yields the highest recognition rates.

| S-EPMC3149884 | biostudies-literature

Domain motion and interdomain hot spots in a multidomain enzyme.

Project description:The aim of this article is to analyze conformational changes by comparing 10 different structures of Pseudomonas aeruginosa phosphomannomutase/phosphoglucomutase (PMM/PGM), a four-domain enzyme in which both substrate binding and catalysis require substantial movement of the C-terminal domain. We focus on changes in interdomain and active site crevices using a method called computational solvent mapping rather than superimposing the structures. The method places molecular probes (i.e., small organic molecules containing various functional groups) around the protein to find hot spots. One of the most important hot spots is in the active site, consistent with the ability of the enzyme to bind both glucose and mannose phosphosugar substrates. The protein has eight additional hot spots at domain-domain interfaces and hinge regions. The locations and nature of six of these hot spots vary between the open, half-open, and closed conformers of the enzyme, in good agreement with the ligand-induced conformational changes. In the closed structures the number of probe clusters at the hinge region significantly depends on the position of the phosphorylated oxygen in the substrate (e.g., glucose 1-phosphate versus glucose 6-phosphate), but the protein remains almost unchanged in terms of the overall RMSD, indicating that computational solvent mapping is a more sensitive approach to detect changes in binding sites and interdomain crevices. Focusing on multidomain proteins we show that the subresolution conformational differences revealed by the mapping are in fact significant, and present a general statistical method of analysis to determine the significance of rigid body domain movements in X-ray structures.

| S-EPMC2975130 | biostudies-literature

Analysis of Multidomain Protein Dynamics.

Project description:Proteins with a modular architecture of multiple domains connected by linkers often exhibit diversity in the relative positions of domains, while the domain tertiary structure remains unchanged. The biological function of these modular proteins, or the regulation of their activity, depends on the variation in domain orientation and separation. Accordingly, careful characterization of interdomain motion and correlated fluctuations of multidomain systems is relevant for understanding the functional behavior of modular proteins. Molecular dynamics (MD) simulations provides a powerful approach to study these motions in atomic detail. Nevertheless, the common procedure for analyzing fluctuations from MD simulations after rigid-body alignment fails for multidomain proteins; it greatly overestimates correlated positional fluctuations in the presence of relative domain motion. We show here that expressing the atomic motions of a multidomain protein as a combination of displacement within the domain reference frame and motion of the relative domains correctly separates the internal motions to allow a useful description of correlated fluctuations. We illustrate the methodology of separating the domain fluctuations and local fluctuations by application to the tandem SH2 domains of human Syk protein kinase and by characterizing an effect of phosphorylation on the dynamics. Correlated motions are assessed from a distance covariance rather than the more common vector-coordinate covariance. The approach makes it possible to calculate the proper correlations in fluctuations internal to a domain as well as between domains.

| S-EPMC4900692 | biostudies-literature

An Empirical Investigation of Domain Generalization with Empirical Risk Minimizers.

Project description:Recent work demonstrates that deep neural networks trained using Empirical Risk Minimization (ERM) can generalize under distribution shift, outperforming specialized training algorithms for domain generalization. The goal of this paper is to further understand this phenomenon. In particular, we study the extent to which the seminal domain adaptation theory of Ben-David et al. (2007) explains the performance of ERMs. Perhaps surprisingly, we find that this theory does not provide a tight explanation of the out-of-domain generalization observed across a large number of ERM models trained on three popular domain generalization datasets. This motivates us to investigate other possible measures-that, however, lack theory-which could explain generalization in this setting. Our investigation reveals that measures relating to the Fisher information, predictive entropy, and maximum mean discrepancy are good predictors of the out-of-distribution generalization of ERM models. We hope that our work helps galvanize the community towards building a better understanding of when deep networks trained with ERM generalize out-of-distribution.

| S-EPMC9806840 | biostudies-literature

ECOD: identification of distant homology among multidomain and transmembrane domain proteins.

Project description:The manual classification of protein domains is approaching its 20th anniversary. ECOD is our mixed manual-automatic domain classification. Over time, the types of proteins which require manual curation has changed. Depositions with complex multidomain and multichain arrangements are commonplace. Transmembrane domains are regularly classified. Repeatedly, domains which are initially believed to be novel are found to have homologous links to existing classified domains. Here we present a brief summary of recent manual curation efforts in ECOD generally combined with specific case studies of transmembrane and multidomain proteins wherein manual curation was useful for discovering new homologous relationships. We present a new taxonomy for the classification of ABC transporter transmembrane domains. We examine alternate topologies of the leucine-specific (LS) domain of Leucine tRNA-synthetase. Finally, we elaborate on a distant homologous links between two helical dimerization domains.

| S-EPMC6588880 | biostudies-literature

Supervised discriminant analysis for droplet micro-magnetofluidics.

Project description:We apply the technique of supervised discriminant analysis (SDA) for in-flow detection in droplet-based magnetofluidics. Based on the SDA, we successfully discriminate bivariant droplets of different volumes containing different encapsulated magnetic content produced by a GMR-based lab-on-chip platform. We demonstrate that the accuracy of discrimination is superior when the correlation of variables for data training is included to the case when the spatial distribution of variables is considered. Droplets produced with differences in ferrofluid concentration of 2.5 mg/ml and volume of 200 pl have been identified with high accuracy (98 %), indicating the significance of SDA for e.g. the discrimination in magnetic immuno-agglutination assays. Furthermore, the results open the way for the development of a unique magnetofluidic platform for future applications in multiplexed droplet-based barcoding assays and screening.

| S-EPMC4562118 | biostudies-literature

Multiblock discriminant analysis for integrative genomic study.

Project description:Human diseases are abnormal medical conditions in which multiple biological components are complicatedly involved. Nevertheless, most contributions of research have been made with a single type of genetic data such as Single Nucleotide Polymorphism (SNP) or Copy Number Variation (CNV). Furthermore, epigenetic modifications and transcriptional regulations have to be considered to fully exploit the knowledge of the complex human diseases as well as the genomic variants. We call the collection of the multiple heterogeneous data "multiblock data." In this paper, we propose a novel Multiblock Discriminant Analysis (MultiDA) method that provides a new integrative genomic model for the multiblock analysis and an efficient algorithm for discriminant analysis. The integrative genomic model is built by exploiting the representative genomic data including SNP, CNV, DNA methylation, and gene expression. The efficient algorithm for the discriminant analysis identifies discriminative factors of the multiblock data. The discriminant analysis is essential to discover biomarkers in computational biology. The performance of the proposed MultiDA was assessed by intensive simulation experiments, where the outstanding performance comparing the related methods was reported. As a target application, we applied MultiDA to human brain data of psychiatric disorders. The findings and gene regulatory network derived from the experiment are discussed.

| S-EPMC4450020 | biostudies-literature

OmicsDI is part of the ELIXIR infrastructure

OmicsDI is an Elixir interoperability service. Learn more ›

Tweets

OmicsDI Databases

PRIDE
PeptideAtlas
MassIVE
JPOST Repository
Physiome Model Repository

EGA
EVA
ENA
LINCS
PAXDB
Cell Collective

MetaboLights
Metabolomics Workbench
MetabolomeExpress
GNPS
BioModels
FAIRDOMHub

ArrayExpress
dbGaP
ExpressionAtlas
GEO
NODE

Information

Databases
Help
API
Contact us
Code on GitHub
Terms of use
Submit Data