Browse
Submit Data
Databases
API
Help

Dataset Information

0 Views

0 Connections

0 Citations

0 Reanalyses

0 Downloads

Omics score: 0

A Contour Stochastic Gradient Langevin Dynamics Algorithm for Simulations of Multi-modal Distributions.

ABSTRACT: We propose an adaptively weighted stochastic gradient Langevin dynamics algorithm (SGLD), so-called contour stochastic gradient Langevin dynamics (CSGLD), for Bayesian learning in big data statistics. The proposed algorithm is essentially a scalable dynamic importance sampler, which automatically flattens the target distribution such that the simulation for a multi-modal distribution can be greatly facilitated. Theoretically, we prove a stability condition and establish the asymptotic convergence of the self-adapting parameter to a unique fixed-point, regardless of the non-convexity of the original energy function; we also present an error analysis for the weighted averaging estimators. Empirically, the CSGLD algorithm is tested on multiple benchmark datasets including CIFAR10 and CIFAR100. The numerical results indicate its superiority over the existing state-of-the-art algorithms in training deep neural networks.

SUBMITTER: Deng W

PROVIDER: S-EPMC8457681 | biostudies-literature | 2020 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Json Xml

Publications

A Contour Stochastic Gradient Langevin Dynamics Algorithm for Simulations of Multi-modal Distributions.

Deng Wei W Lin Guang G Liang Faming F

Advances in neural information processing systems 20201201

We propose an adaptively weighted stochastic gradient Langevin dynamics algorithm (SGLD), so-called contour stochastic gradient Langevin dynamics (CSGLD), for Bayesian learning in big data statistics. The proposed algorithm is essentially a <i>scalable dynamic importance sampler</i>, which automatically <i>flattens</i> the target distribution such that the simulation for a multi-modal distribution can be greatly facilitated. Theoretically, we prove a stability condition and establish the asympto ...[more]

PMID: 34556969

Similar Datasets

Variance Reduction in Stochastic Gradient Langevin Dynamics.

Project description:Stochastic gradient-based Monte Carlo methods such as stochastic gradient Langevin dynamics are useful tools for posterior inference on large scale datasets in many machine learning applications. These methods scale to large datasets by using noisy gradients calculated using a mini-batch or subset of the dataset. However, the high variance inherent in these noisy gradients degrades performance and leads to slower mixing. In this paper, we present techniques for reducing variance in stochastic gradient Langevin dynamics, yielding novel stochastic Monte Carlo methods that improve performance by reducing the variance in the stochastic gradient. We show that our proposed method has better theoretical guarantees on convergence rate than stochastic Langevin dynamics. This is complemented by impressive empirical results obtained on a variety of real world datasets, and on four different machine learning tasks (regression, classification, independent component analysis and mixture modeling). These theoretical and empirical contributions combine to make a compelling case for using variance reduction in stochastic Monte Carlo methods.

| S-EPMC5508544 | biostudies-literature

Learning non-stationary Langevin dynamics from stochastic observations of latent trajectories.

Project description:Many complex systems operating far from the equilibrium exhibit stochastic dynamics that can be described by a Langevin equation. Inferring Langevin equations from data can reveal how transient dynamics of such systems give rise to their function. However, dynamics are often inaccessible directly and can be only gleaned through a stochastic observation process, which makes the inference challenging. Here we present a non-parametric framework for inferring the Langevin equation, which explicitly models the stochastic observation process and non-stationary latent dynamics. The framework accounts for the non-equilibrium initial and final states of the observed system and for the possibility that the system's dynamics define the duration of observations. Omitting any of these non-stationary components results in incorrect inference, in which erroneous features arise in the dynamics due to non-stationary data distribution. We illustrate the framework using models of neural dynamics underlying decision making in the brain.

| S-EPMC8514604 | biostudies-literature

Many-particle Brownian and Langevin Dynamics Simulations with the Brownmove package.

Project description:BackgroundBrownian Dynamics (BD) is a coarse-grained implicit-solvent simulation method that is routinely used to investigate binary protein association dynamics, but due to its efficiency in handling large simulation volumes and particle numbers it is well suited to also describe many-protein scenarios as they often occur in biological cells.ResultsHere we introduce our "brownmove" simulation package which was designed to handle many-particle problems with varying particle numbers and allows for a very flexible definition of rigid and flexible protein and polymer models. Both a Brownian and a Langevin dynamics (LD) propagation scheme can be used and hydrodynamic interactions are treated efficiently with our recently introduced TEA-HI ansatz [Geyer, Winter, JCP 130 (2009) 114905]. With simulations of constrained polymers and flexible models of spherical proteins we demonstrate that it is crucial to include hydrodynamics when multi-bead models are used in BD or LD simulations. Only then both the translational and the rotational diffusion coefficients and the timescales of the internal dynamics can be reproduced correctly. In the third example project we show how constant density boundary conditions [Geyer et al, JCP 120 (2004) 4573] can be used to set up a non-equilibrium simulation of diffusional transport across an array of fixed obstacles. Finally, we demonstrate how the agglomeration dynamics of multiple particles with attractive patches can be analysed conveniently with the help of a dynamic interaction network.ConclusionsCombining BD and LD propagation, fast hydrodynamics, a flexible protein model, and interfaces for "open" simulation settings, our freely available "brownmove" simulation package constitutes a new platform for coarse-grained many-particle simulations of biologically relevant diffusion and transport processes.

| S-EPMC3098214 | biostudies-literature

A Stochastic Multiscale Model of Cardiac Thin Filament Activation Using Brownian-Langevin Dynamics.

Project description:We use Brownian-Langevin dynamics principles to derive a coarse-graining multiscale myofilament model that can describe the thin-filament activation process during contraction. The model links atomistic molecular simulations of protein-protein interactions in the thin-filament regulatory unit to sarcomere-level activation dynamics. We first calculate the molecular interaction energy between tropomyosin and actin surface using Brownian dynamics simulations. This energy profile is then generalized to account for the observed tropomyosin transitions between its regulatory stable states. The generalized energy landscape then served as a basis for developing a filament-scale model using Langevin dynamics. This integrated analysis, spanning molecular to thin-filament scales, is capable of tracking the events of the tropomyosin conformational changes as it moves over the actin surface. The tropomyosin coil with flexible overlap regions between adjacent tropomyosins is represented in the model as a system of coupled stochastic ordinary differential equations. The proposed multiscale approach provides a more detailed molecular connection between tropomyosin dynamics, the trompomyosin-actin interaction-energy landscape, and the generated force by the sarcomere.

| S-EPMC6990154 | biostudies-literature

PIK3CA somatic mutations in breast cancer: Mechanistic insights from Langevin dynamics simulations.

Project description:Somatic mutations in PIK3CA (phosphatidylinositol-3 kinase, catalytic subunit, alpha isoform) are reported in breast and other human cancers to concentrate at hotspots within its kinase and helical domains. Most of these mutations cause kinase gain of function in vitro and are associated with oncogenicity in vivo. However, little is known about the mechanisms driving tumor development. We have performed computational structural studies on a homology model of wildtype PIK3CA plus recurrent H1047R, H1047L, and P539R mutations, located in the kinase and helical domains, respectively. The time evolution of the structures show that H1047R/L mutants exhibit a larger area of the catalytic cleft between the kinase N- and C-lobes compared with the wildtype that could facilitate the entrance of substrates. This larger area might yield enhanced substrate-to-product turnover associated with oncogenicity. In addition, the H1047R/L mutants display increased kinase activation loop mobility, compared with the wildtype. The P539R mutant forms more hydrogen bonds and salt-bridge interactions than the wildtype, properties that are associated with enhanced thermostability. Mutant-specific differences in the catalytic cleft and activation loop behavior suggest that structure-based mutant-specific inhibitors can be designed for PIK3CA-positive breast cancers.

| S-EPMC3094910 | biostudies-literature

Multi-omics data and stochastic simulations support a cancer heterogeneity framework unifying genetic, epigenetic, and stochastic variabilities

Project description:This SuperSeries is composed of the SubSeries listed below.

2020-05-08 | GSE150084 | GEO

Multi-omics data and stochastic simulations support a cancer heterogeneity framework unifying genetic, epigenetic, and stochastic variabilities [scRNA-seq]

Project description:We subject multiple versions and single cell-derived sublines of an archetypal non-small cell lung cancer cell line to experimental scrutiny at the genomic, transcriptomic, and cell population levels. We find evidence of genetic variability among cell line versions and one subline. Additional sublines show no genetic variation but epigenetic differences are detected at the transcriptomic level. Stochastic simulations of subline growth dynamics confirm that one subline likely contains at least two genetic states while all others are isogenic. Overall, our results substantiate a view of intratumoral heterogeneity in which different genetic states give rise to multiple epigenetic “basins of attraction,” across which cells can transition driven by stochastic factors such as gene expression noise and asymmetric cell division. Deconvolving intratumoral heterogeneity into genetic, epigenetic, and stochastic components provides a lens through which to view tumor drug response and acquired treatment resistance that may lead to novel therapeutic strategies in the future.

2020-05-08 | GSE150083 | GEO

Fast stochastic algorithm for simulating evolutionary population dynamics.

Project description:MotivationMany important aspects of evolutionary dynamics can only be addressed through simulations. However, accurate simulations of realistically large populations over long periods of time needed for evolution to proceed are computationally expensive. Mutants can be present in very small numbers and yet (if they are more fit than others) be the key part of the evolutionary process. This leads to significant stochasticity that needs to be accounted for. Different evolutionary events occur at very different time scales: mutations are typically much rarer than reproduction and deaths.ResultsWe introduce a new exact algorithm for fast fully stochastic simulations of evolutionary dynamics that include birth, death and mutation events. It produces a significant speedup compared to direct stochastic simulations in a typical case when the population size is large and the mutation rates are much smaller than birth and death rates. The algorithm performance is illustrated by several examples that include evolution on a smooth and rugged fitness landscape. We also show how this algorithm can be adapted for approximate simulations of more complex evolutionary problems and illustrate it by simulations of a stochastic competitive growth model.

| S-EPMC3338019 | biostudies-literature

Multi-omics data and stochastic simulations support a cancer heterogeneity framework unifying genetic, epigenetic, and stochastic variabilities [Bulk-RNA-seq]

2020-05-08 | GSE150025 | GEO

Stochastic simulations of the origins and implications of long-tailed distributions in gene expression.

Project description:Gene expression noise results in protein number distributions ranging from long-tailed to Gaussian. We show how long-tailed distributions arise from a stochastic model of the constituent chemical reactions and suggest that, in conjunction with cooperative switches, they lead to more sensitive selection of a subpopulation of cells with high protein number than is possible with Gaussian distributions. Single-cell-tracking experiments are presented to validate some of the assumptions of the stochastic simulations. We also examine the effect of DNA looping on the shape of protein distributions. We further show that when switches are incorporated in the regulation of a gene via a feedback loop, the distributions can become bimodal. This might explain the bimodal distribution of certain morphogens during early embryogenesis.

| S-EPMC555697 | 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