This work proposes a mathematical framework for loss landscapes and optimization in deep neural networks.
problem The effectiveness of gradient-based optimization in over-parameterized neural networks.
method A modern view and mathematical framework of loss landscapes and efficient optimization in over-parameterized machine learning models.
result Wide neural networks satisfy the PL∗ condition, explaining (S)GD convergence to a global minimum. CGNS predicts probabilistic trajectories for safer autonomous systems.
problem Accurate probabilistic trajectory prediction for dynamic obstacles in complex scenarios.
method CGNS combines latent space learning and variational divergence minimization, incorporating static and interaction information with soft attention mechanisms and regularization for soft constraints.
result CGNS outperforms baseline approaches in pedestrian trajectory prediction and naturalistic driving datasets.
Develops CLDS models to model neural activity with nonlinear dynamics.
problem Complex, nonlinear dynamics in neural population activity.
method Conditionally Linear Dynamical System (CLDS) models using Gaussian Process (GP) priors.
result CLDS models can perform well even in data-limited conditions.
Paper develops neural network for Mandarin polyphone disambiguation.
problem Homograph problem in Mandarin Chinese text-to-speech.
method Bidirectional RNN for context, prediction network for mapping embeddings to pronunciations.
result Achieves 94.69% accuracy on polyphonic character dataset.
Study shows neural operators can efficiently solve complex reaction-diffusion systems.
problem Efficiently solving nonlinear reaction-diffusion systems using neural operators.
method Laplacian-based neural operators applied to a generalized Gierer-Meinhardt system.
result Explicit approximation error bounds established for neural operators in terms of network parameters.
Recently a variety of LSTM-based conditional language models (LM) have been applied across a range of language generation tasks. In this work we study various model architectures and different ways to represent and aggregate the source information in an end-to-end neural dialogue system framework. A method called snaps…
Enhanced model predicts chaotic systems with improved long-term accuracy.
problem Learning chaotic systems and long-term predictions from incomplete data.
method Path-dependent Neural Jump ODE (PD-NJ-ODE) model for online prediction.
result The model matches true chaotic system dynamics closely and improves long-term predictions.
New method uses Gaussian processes for solving linear PDEs with boundary conditions.
problem Solving linear PDEs with boundary conditions.
method Boundary Ehrenpreis--Palamodov Gaussian Processes (B-EPGPs).
result Significant accuracy and resource improvements over existing methods.
Study approximates operators on labelled conditional distributions for non-exchangeable systems.
problem Approximating operators on constrained probability measures for non-exchangeable systems.
method Combines cylindrical approximations and DeepONet-type neural architecture for finite-dimensional representations.
result Establishes a universal approximation theorem for continuous operators on Mλ. Modern automatic speech recognition (ASR) systems need to be robust under acoustic variability arising from environmental, speaker, channel, and recording conditions. Ensuring such robustness to variability is a challenge in modern day neural network-based ASR systems, especially when all types of variability are not s…
Efficiently detects failures in autonomous systems using adaptive stress testing.
problem Identifying likely failure scenarios in autonomous systems.
method Employed a recurrent neural network to simulate initial conditions and detect failures.
result Solver can now find solutions for previously intractable problems.
Research proposes a test case generation system for deep learning models using dataset properties.
problem Automated generation of extensive test cases for deep learning models is challenging.
method Measures dataset quality and proposes a test case generation system guided by dataset properties.
result Systematic test case generation for deep learning models is effective.
FNSDA adapts to new dynamics via Fourier space adaptation.
problem Generalizing to unseen dynamical systems with limited data.
method Automatic partitioning of known environments in Fourier modes and adaptation of specific modes for new environments.
result FNSDA achieves superior or competitive generalization performance with reduced parameter cost.
Study on neural scaling laws for solving linear systems in-context.
problem Theoretical guarantees for solving linear systems using a linear transformer architecture.
method Neural scaling laws and task diversity for in-domain and out-of-domain generalization.
result Novel notion of task diversity for necessary and sufficient condition of generalization under task shifts.
Generative model uses neural flows for next-frame video generation conditioned on labels.
problem Blurriness and instability in video generation models.
method Proposes using Glow, a neural flow model, for next-frame video generation conditioned on labels.
result Glow model produces clearer and more stable videos compared to GANs.
AI-driven framework optimizes MCMC-based preconditioners for faster linear system solving.
problem Slow convergence of Krylov subspace solvers for ill-conditioned matrices.
method Graph neural surrogate and Bayesian optimization for AI-tuned MCMC parameters.
result 50% reduction in iterations to convergence on unseen system.
New method models unknown systems with hidden parameters using neural networks.
problem Modeling unknown dynamical systems with hidden parameters.
method Training a deep neural network (DNN) model using trajectory data of the unknown system.
result DNN model accurately predicts unknown dynamical systems with new initial conditions.
GFM models neural network training as a dynamical system to forecast final weights.
problem Computational intensity and inefficiency in training deep neural networks.
method Gradient Flow Matching (GFM) treats training as a dynamical system with learned vector fields.
result GFM achieves forecasting accuracy competitive with Transformer-based models and significantly outperforms classical baselines.
New metric predicts neural network reliability under novel conditions.
problem Verifying neural networks' safety in novel scenarios.
method ML Dependability metric, Task Undependability, Harmful Undependability.
result Accurately predicts reliability under novel conditions.
This paper shows overparameterized deep neural networks are convex and learn useful features.
problem Analyzing fully trained overparameterized deep neural networks.
method Generalized neural feature repopulation technique.
result Overparameterized deep neural networks are inherently convex and learn useful features.
New method generates equilibrium glass configurations efficiently.
problem Sampling equilibrium configurations of amorphous materials is slow and difficult.
method Riemannian stochastic interpolation framework combining Riemannian stochastic interpolant and equivariant flow matching.
result Enforcing geometric and symmetry constraints significantly improves generative performance.
Enhances neural network solvers for PDEs with complex boundary conditions.
problem Challenges in solving PDEs with high accuracy and complex boundary conditions.
method Integrates natural gradient optimization with numerical time-stepping schemes to enforce Dirichlet boundary conditions.
result Superior accuracy and computational efficiency of the proposed methods for solving PDEs.
Deep neural nets approximate random dynamical system trajectories uniformly in time.
problem Approximating trajectories of random dynamical systems over infinite time horizons.
method Recurrent neural networks with simple feedback structures.
result Certain random trajectories can be approximated uniformly in time to any desired accuracy.
Physics-enhanced NNs improve predictive accuracy in small data scenarios.
problem Predicting physical systems dynamics with limited data.
method Integrating physical principles (Hamiltonian/Lagrangian) and regularization term based on energy level into neural networks.
result Significant gains in predictive accuracy for small data cases.
Neuroscience is experiencing a data revolution in which many hundreds or thousands of neurons are recorded simultaneously. Currently, there is little consensus on how such data should be analyzed. Here we introduce LFADS (Latent Factor Analysis via Dynamical Systems), a method to infer latent dynamics from simultaneous…
Generative adversarial network for probabilistic forecasting of random systems.
problem Forecasting random dynamical systems without distributional assumptions.
method Recurrent neural network and generative adversarial network (GAN) with regularization based on maximum mean discrepancy (MMD).
result The proposed model successfully forecasts complex stochastic processes with multiple-step predictions.
Personal VAD detects target speaker voice activity efficiently.
problem Efficiently detect target speaker voice activity for reduced computational cost and battery usage.
method Trains a neural network conditioned on speaker embedding or verification score, outputs probabilities for three speech classes.
result Trained model with 130K parameters outperforms combined standard VAD and speaker recognition networks.
Investigates neural TTS systems for Japanese and English.
problem Improving neural TTS systems for high-quality speech synthesis.
method Comparative study of neural sequence-to-sequence TTS vs. DNN pipeline TTS, varying model architecture, parameter size, and language.
result A neural sequence-to-sequence TTS system requires sufficient model parameters and a powerful encoder for high-quality speech synthesis.
Method discovers local independence in systems with continuous variables.
problem Applying Context-Specific Independence (CSI) to continuous variables is impractical.
method Neural contextual decomposition (NCD) learns partition of joint outcome space.
result NCD successfully discovers local independence in synthetic and real-world systems.
Discovery of atomistic systems with desirable properties is a major challenge in chemistry and material science. Here we introduce a novel, autoregressive, convolutional deep neural network architecture that generates molecular equilibrium structures by sequentially placing atoms in three-dimensional space. The model e…
The large thermal capacity of buildings enables heating, ventilating, and air-conditioning (HVAC) systems to be exploited as demand response (DR) resources. Optimal DR of HVAC units is challenging, particularly for multi-zone buildings, because this requires detailed physics-based models of zonal temperature variations…
Symbolic regression is a type of discrete optimization problem that involves searching expressions that fit given data points. In many cases, other mathematical constraints about the unknown expression not only provide more information beyond just values at some inputs, but also effectively constrain the search space. …
Paper introduces non-adversarial training for Neural SDEs using signature kernel scores.
problem Stability and mode collapse issues in adversarial training of Neural SDEs.
method Uses signature kernel scores as objective function for non-adversarial training.
result Non-adversarial training leads to better performance and more stable models.
Neural surrogate predicts SPN rates from token trajectories.
problem Challenging parameter estimation in SPNs with covariates.
method 1D Convolutional Residual Network trained on Gillespie-simulated SPN realizations.
result Surrogate predicts rate-function coefficients with RMSE = 0.043.
Estimates causal effects using neural autoregressive density estimators.
problem Estimating causal effects in non-linear systems.
method Neural autoregressive density estimators within Pearl's do-calculus framework.
result Retrieves causal effects from non-linear systems without explicit modeling.
Equivariant neural networks improve performance and generalization in lattice field theory tasks.
problem Improving neural network performance and generalization in lattice field theory.
method Investigation of translationally equivariant neural networks in a two-dimensional scalar field model.
result Equivariant neural networks significantly outperform non-equivariant ones in various tasks, including physical parameters and lattice sizes.
VerifAI toolkit improves neural network-based aircraft taxiing system safety.
problem Improving safety of autonomous aircraft taxiing systems using neural networks.
method Unified approach to formal analysis and retraining of AI systems, including falsification, debugging, and retraining.
result Improved neural network performance and reduced failure cases in aircraft taxiing system.
A new hierarchy quantifies agency in systems based on information processing.
problem Lack of a measurable, universal definition for agency in intelligent systems.
method Developed a bottom-up framework based on information processing hierarchy.
result Identified three orders of information processing (I, II, III) as necessary for agency.
NeuralIF uses neural networks to improve preconditioning for faster CG convergence.
problem Improving convergence of conjugate gradient method for large-scale sparse systems.
method Data-driven approach using graph neural networks to generate incomplete factorization.
result Data-driven preconditioners accelerate convergence of conjugate gradient method.
DGNet solves complex dynamical systems with neural networks and constraints.
problem Real-time accurate solutions for large-scale complex systems.
method Model-constrained discontinuous Galerkin Network (DGNet) for compressible Euler equations.
result DGNet achieves out-of-distribution generalization and improved stability.
Paper proposes hybrid machine learning for tuning first principles models in engineering systems.
problem Inaccurate first principles models in process engineering due to changing conditions.
method Hybrid machine learning framework using Bayesian Neural Networks.
result Uncertainty estimates improve operation decisions in multiphase flow modeling.
Discover conservation laws from trajectories using a neural network.
problem Finding invariants and conservation laws from large-scale data without prior knowledge.
method ConservNet, a neural network trained with noise-variance loss to discover hidden invariants in grouped multi-dimensional observables.
result Successfully discovers underlying invariants from simulated and real-world systems.
Proposes a new method for data assimilation using closed-form conditional diffusion models.
problem Data assimilation for systems with complex, non-Gaussian probability distributions.
method Uses kernel density estimation to model joint distributions and leverages the score function for efficient evaluation.
result Outperforms ensemble Kalman and particle filters in nonlinear data assimilation problems.
Recurrent Neural Processes model time series with conditional independence to capture slow variabilities efficiently.
problem Modeling time series data with slow long-term variabilities efficiently.
method Recurrent Neural Processes (RNP) model state space with conditional independence among subsequences.
result RNP state spaces improve predictive performance on real-world time-series data and nonlinear system identification.
Paper certifies neural network control policies against persistent adversarial perturbations.
problem Neural networks' fragility to adversarial perturbations in control systems.
method Combining neural network certification tools with robust control theory.
result Certifies neural network policies in a control loop under l-infinity norm bounded adversarial perturbations.
Tensor completion method identifies nonlinear systems from input-output data.
problem Identifying nonlinear functions from input-output data pairs.
method Formulated as tensor completion problem with smoothness regularization and solved using block coordinate descent.
result Provable correct nonlinear system identification under certain conditions.
A GAN method for stochastic boundary conditions in fast dynamics.
problem Uncertainty quantification in fast dynamics and wave propagation.
method Physics-informed GANs for stochastic boundary conditions.
result Improved convergence and better stochastic boundary conditions.
Proposes a new framework for open set recognition using conditional probabilistic generative models.
problem Unknown samples can mislead traditional deep neural networks during testing.
method Conditional Probabilistic Generative Models (CPGM) that combine generative models with discriminative information.
result Significantly outperforms baselines on multiple benchmark datasets.