MO-GP models fill gaps in biophysical data with across-domain info transfer.
problem Gap filling of biophysical parameters LAI and fAPAR over rice areas.
method Multi-output Gaussian Processes (MO-GP) based on Linear Model of Coregionalization (LMC).
result MO-GP models successfully predict biophysical variables even in high missing data regimes.
Paper introduces new regression methods for consistent estimation of biophysical parameters.
problem Estimating biophysical parameters while respecting auxiliary variables.
method Linear and nonlinear kernel-based regression models with consistency constraints.
result Models provide closed-form solutions and successfully estimate chlorophyll content.
PINNs solve neuronal parameter and state estimation problems with limited data.
problem Estimating parameters and hidden state variables from noisy partial data in multiscale neuronal models.
method Physics-informed neural networks (PINNs) for joint state and parameter estimation.
result PINNs deliver robust and accurate parameter inference and state reconstruction, even with limited data.
Joint Gaussian Processes combine real and simulated data for better biophysical parameter retrieval.
problem Inverting radiative transfer models for accurate biophysical parameter estimation.
method Joint Gaussian Process (JGP) that combines real and simulated data for regression.
result JGP outperforms traditional methods in leaf area index retrieval from Landsat data.
Biophysical models explain deep learning in gene regulation.
problem Difficulty in interpreting deep learning models in gene regulation.
method Expressed biophysical models as neural networks with explicit interpretations.
result Biophysical networks can be inferred from MPRAs.
Hybrid Amortized Inference improves PPG model interpretability.
problem Tension between PPG biomarker accuracy and clinical interpretability.
method Introduces PPGen for biophysical PPG signal-physiological parameter relation, and HAI for fast, robust estimation.
result Hybrid Amortized Inference accurately infers physiological parameters from PPG signals.
PS-VAE extracts multi-parameter MRI biomarkers with uncertainty quantification.
problem Uncertainty in inverse problems limits clinical acceptance of quantitative MRI methods.
method Physics-Structured Variational Autoencoder (PS-VAE) integrating physics simulator and self-supervised learning.
result PS-VAE provides full covariance of inter-parameter correlations and accelerates multi-parametric MRI quantification.
Novel framework for medical image segmentation using deep learning.
problem Class imbalance and domain adaptation in medical image segmentation.
method Biophysics-based domain adaptation and automatic segmentation of white, gray, and cerebrospinal fluid.
result Improved segmentation performance, especially with the biophysics-based domain adaptation.
Deep learning predicts tissue properties from cell-laden hydrogels.
problem Predicting tissue properties from cell-laden hydrogels.
method Developed a process for generating mould designs, created a training set of 6500 cases, trained a deep learning model (pix2pix).
result Deep learning makes excellent predictions and is significantly faster than biophysical methods.
Time-lagged VAE reduces complex dynamics to a single embedding.
problem Interpreting high-dimensional time-series data for nonlinear systems.
method Variational dynamics encoder (VDE) using time-lagged variational autoencoders.
result Captures nontrivial dynamics in various examples, including protein folding.
Generative model learns conditional distributions on collective variable levels.
problem Modeling conditional probability distributions on collective variable levels.
method General and efficient learning approach, data enrichment strategy.
result Effective generative models on different level-sets of collective variables.
Despite our extensive knowledge of biophysical properties of neurons, there is no commonly accepted algorithmic theory of neuronal function. Here we explore the hypothesis that single-layer neuronal networks perform online symmetric nonnegative matrix factorization (SNMF) of the similarity matrix of the streamed data. …
GP CaKe models causal brain connectivity using Gaussian processes.
problem Understanding how one brain region drives activity in another.
method Integro-differential equations and causal kernels learned via Gaussian process regression.
result Demonstrated efficacy on simulations and MEG data.
Biologically inspired neural networks resist adversarial attacks.
problem Resisting adversarial attacks on deep neural networks.
method Developed a scheme based on biophysical principles of neural circuits to train deep networks to be robust to adversarial attacks.
result Generated deep neural networks achieve state-of-the-art performance on adversarial examples without exposure during training.
Synthesizes computational approaches to understand neural timescales.
problem Varying definitions and measurements of neural timescales across studies.
method Reviews data analysis methods, biophysical models, and machine learning models.
result Complements experimental studies with a holistic view of neural timescales.
Generative models accelerate molecular dynamics by four orders of magnitude.
problem Femtosecond time steps limit access to slow molecular processes.
method Deep generative modeling framework that accelerates sampling.
result Quantitative characterization of equilibrium ensembles and dynamical relaxation processes.
Estimates tumor growth parameters from MRI data.
problem Personalizing brain tumor growth models for individual patients.
method Learning-based technique using a mixture-density network to estimate parameters from medical scans.
result Reduces the number of forward model integrations and relaxes functional form constraints.
A neuron is a basic physiological and computational unit of the brain. While much is known about the physiological properties of a neuron, its computational role is poorly understood. Here we propose to view a neuron as a signal processing device that represents the incoming streaming data matrix as a sparse vector of …
PRISM infers model structures and parameters from simulations, controlling complexity at test time.
problem Choosing among large model families for scientific discovery.
method Simulation-based encoder-decoder that infers model structures and parameters, with test-time complexity control.
result PRISM scales to large model families and performs model selection in biophysical diffusion MRI.
New ABC method for Bayesian inference of neural models.
problem Challenging to identify which mechanistic models quantitatively reproduce neural data.
method Likelihood-free inference using neural networks and Bayesian mixture-density networks.
result Efficiently estimates posterior distributions and recovers ground-truth parameters.
New ABC method uses neural emulators for inference in complex models.
problem Approximate Bayesian Computation (ABC) for models without tractable likelihoods.
method Probabilistic neural emulator networks to learn synthetic likelihoods, adaptive simulations.
result Accurate and efficient inference on high-dimensional problems.
Multi-StyleGAN simulates live cell microscopy imagery.
problem Costly and complex live cell experiments.
method Generative adversarial network (GAN) synthesizing multi-domain time-lapse images.
result Captures biophysical factors and time dependencies in cell imagery.
Mathematician summarizes protein geometry and mutation effects.
problem Understanding how proteins mutate and their structure-function relationship.
method Mathematical analysis of protein structures and functions, focusing on hydrogen bonds and secondary structure.
result Protein secondary structure regulates mutation by stabilizing or destabilizing regions.
Learning and memory in the brain are implemented by complex, time-varying changes in neural circuitry. The computational rules according to which synaptic weights change over time are the subject of much research, and are not precisely understood. Until recently, limitations in experimental methods have made it challen…
A new method for analyzing shapes and forms using additive models on manifolds.
problem Analyzing shapes and forms under geometric transformations.
method Extending generalized additive regression to models for shapes/forms using squared geodesic distance and Riemannian L2-Boosting algorithm. result Automated model selection and intuitive visualization of covariate effects in shape/form space.
Novel parallel GNN predicts protein-ligand interactions with high accuracy.
problem Accurate prediction of protein-ligand interactions for drug design.
method Parallel Graph Neural Networks (GNN) integrating 3D structural data.
result GNN achieves high accuracy in predicting binary interactions and activity.
Model shows consumption growth slows due to finite planet resources.
problem Understanding and predicting consumption growth in a finite planet context.
method Logistic model of consumption growth, cumulant expansion method.
result Social discount rates decline over time due to planetary resource constraints.
SIM-CE models C. elegans neural circuits for behavioral analysis.
problem Understanding the neural basis of C. elegans behavior.
method User-friendly Simulink platform with detailed neuron and synapse models.
result SIM-CE enables detailed multi-scale simulations of C. elegans behavior.
Boosting algorithms predict financial vulnerability of farmers in Chile and Tunisia.
problem Predict financial vulnerability of farmers in Chile and Tunisia using environmental data.
method Interpretable boosting algorithms based on ridge-regularized generalized linear models.
result Interaction effects improve predictive power only when included in two-step boosting.
Extends SINDy to model stochastic dynamical systems.
problem Modeling stochastic dynamical systems from data.
method Sparse Identification of Nonlinear Dynamics (SINDy) extended to stochastic systems, with proof of asymptotic correctness and practical algorithms.
result Proves asymptotic correctness of stochastics SINDy in infinite data limit.
We present a machine learning framework for modeling protein dynamics. Our approach uses L1-regularized, reversible hidden Markov models to understand large protein datasets generated via molecular dynamics simulations. Our model is motivated by three design principles: (1) the requirement of massive scalability; (2) t…
Method reconstructs neuron models from spike times efficiently.
problem Reconstructing neuron models from spike times in degenerate populations.
method Combining deep learning with DICs to map spike times to DIC densities and generate degenerate CBM populations.
result Fast and scalable reconstruction of degenerate populations from spike recordings.
New method infers diffusion equations from sparse data.
problem Statistical inference of diffusion equations from limited data.
method Neural network-based estimators for drift and diffusion tensor.
result Statistical convergence guarantees for Hölder continuous processes.
Pipeline learns topological features for protein stability prediction.
problem Predicting protein stability using topological features.
method Data-driven method to learn topological features, comparing with expert features.
result Topological features achieve 92%-99% of SME-based models' performance.
Lipid necks, large curvature bridges, are shown to be metastable.
problem Understanding the energetically prohibitive yet ubiquitous lipid necks in cell membranes.
method Geometric triality approach to demonstrate metastability.
result Lipid necks can exist for finite but potentially long times without stabilizing mechanisms.
Enhances neural networks with prior function values to improve accuracy.
problem Improving neural network accuracy in regions without training data.
method Develops a probabilistic approach to augment BNNs with prior function values.
result Predictions rely more on prior information in uncertain regions.
MoleculeNet benchmarks molecular machine learning algorithms.
problem Lack of a standard benchmark for molecular machine learning.
method Curated multiple public datasets, established evaluation metrics, released open-source implementations.
result Learnable representations offer the best performance in molecular machine learning.
Machine learning detects subtle glucose changes for early diabetes diagnosis.
problem Challenging early-stage diabetes diagnosis due to subtle glucose changes.
method Applied machine learning to synthetic glucose profiles generated by a biophysical model.
result High accuracy (above 85%) in detecting insulin resistance using various neural networks.
When governed by underlying low-dimensional dynamics, the interdependence of simultaneously recorded population of neurons can be explained by a small number of shared factors, or a low-dimensional trajectory. Recovering these latent trajectories, particularly from single-trial population recordings, may help us unders…
Transfer neural networks for efficient protein dynamics sampling.
problem Efficiently sampling protein dynamics in related systems.
method Variational auto-encoder framework with latent embedding for collective variable.
result Transferable model trained on one protein can efficiently sample related mutants.
Deep learning predicts protein structures accurately.
problem Predicting the 3D structure of proteins from amino acid sequences.
method Embeddings and deep learning models for backbone atom distance matrices and torsion angles.
result Competitive results in CASP13 and CASP12, surpassing previous winners.
Deep learning method segments LA and PPVs from MRI for cardiac disease analysis.
problem Quantitative analysis of left atrium and proximal pulmonary veins from MRI.
method Multi-view CNN with adaptive fusion and loss function for efficient and accurate segmentation.
result Proposed method achieved state-of-the-art sensitivity, specificity, and precision in cardiac segmentation challenge.
Solves the initial CV problem for molecular simulations using machine learning.
problem Selecting appropriate collective variables for enhancing sampling in molecular simulations.
method Data-driven approach inspired by supervised machine learning (SML).
result Various SML algorithms can be used as initial collective variables (SML_cv) for accelerated sampling.
Deep learning predicts preterm birth risk with improved accuracy.
problem Improving accuracy in predicting spontaneous preterm deliveries.
method U-Net segmentation network for automatic extraction of cervical length and anterior cervical angle.
result Combined markers reduce false-negative ratio from 30% to 18%
POSCMs extend SCMs for causal modeling with latent contexts.
problem Causal modeling with latent contexts and endogenous mechanisms.
method Kolmogorov-Arnold-Sprecher edge-functional decomposition for explicit parametrization.
result Identifiability of structure and mechanisms under latent context.
Machine discovers PDEs from spatiotemporal data without prior knowledge.
problem Discovering PDEs from complex spatiotemporal data without prior knowledge.
method Sparse Spatiotemporal System Discovery (extS3extd) using Sparse Bayesian Learning. result Automatically discovers ten types of PDEs from simulation data.
Deep CNN models simulate cognitive deficits from neurodegenerative diseases and TBI.
problem Limited ability to assess damaged neurons in vivo for accurate diagnosis and prognosis.
method Used convolutional neural networks (CNNs) to damage simulated brain connections based on biophysically relevant data on FAS.
result Damage to simulated brain connections leads to human-like cognitive mistakes and quantifiable accuracy reductions.
Proposes a new method to better understand complex system interactions.
problem Current methods like Granger causality and transfer entropy fail to capture higher-order interactions.
method Introduces a generalized approach to capture multivariate causal interactions.
result The method can distinguish causal roles in synergetic interactions.