Sampling weights of deep neural networks

EL Bolager, I Burak, C Datar, Q Sun… - Advances in Neural …, 2023 - proceedings.neurips.cc
We introduce a probability distribution, combined with an efficient sampling algorithm, for
weights and biases of fully-connected neural networks. In a supervised learning context, no …

Data-efficient operator learning via unsupervised pretraining and in-context learning

W Chen, J Song, P Ren… - Advances in …, 2025 - proceedings.neurips.cc
Recent years have witnessed the promise of coupling machine learning methods and
physical domain-specific insights for solving scientific problems based on partial differential …

Learning integral operators via neural integral equations

E Zappala, AHO Fonseca, JO Caro… - Nature Machine …, 2024 - nature.com
Nonlinear operators with long-distance spatiotemporal dependencies are fundamental in
modelling complex systems across sciences; yet, learning these non-local operators …

Group equivariant fourier neural operators for partial differential equations

J Helwig, X Zhang, C Fu, J Kurtin… - arxiv preprint arxiv …, 2023 - arxiv.org
We consider solving partial differential equations (PDEs) with Fourier neural operators
(FNOs), which operate in the frequency domain. Since the laws of physics do not depend on …

Neural spectral methods: Self-supervised learning in the spectral domain

Y Du, N Chalapathi, A Krishnapriyan - arxiv preprint arxiv:2312.05225, 2023 - arxiv.org
We present Neural Spectral Methods, a technique to solve parametric Partial Differential
Equations (PDEs), grounded in classical spectral methods. Our method uses orthogonal …

Memory-less scattering imaging with ultrafast convolutional optical neural networks

Y Zhang, Q Zhang, H Yu, Y Zhang, H Luan, M Gu - Science Advances, 2024 - science.org
The optical memory effect in complex scattering media including turbid tissue and speckle
layers has been a critical foundation for macroscopic and microscopic imaging methods …

How Expressive are Spectral-Temporal Graph Neural Networks for Time Series Forecasting?

M **, G Shi, YF Li, Q Wen, B **ong, T Zhou… - arxiv preprint arxiv …, 2023 - arxiv.org
Spectral-temporal graph neural network is a promising abstraction underlying most time
series forecasting models that are based on graph neural networks (GNNs). However, more …

Neural integral equations

E Zappala, AHO Fonseca, JO Caro, AH Moberly… - arxiv preprint arxiv …, 2022 - arxiv.org
Nonlinear operators with long distance spatiotemporal dependencies are fundamental in
modeling complex systems across sciences, yet learning these nonlocal operators remains …

Learning space-time continuous latent neural pdes from partially observed states

V Iakovlev, M Heinonen… - Advances in Neural …, 2023 - proceedings.neurips.cc
We introduce a novel grid-independent model for learning partial differential equations
(PDEs) from noisy and partial observations on irregular spatiotemporal grids. We propose a …

Understanding the Expressivity and Trainability of Fourier Neural Operator: A Mean-Field Perspective

T Koshizuka, M Fujisawa… - Advances in Neural …, 2025 - proceedings.neurips.cc
In this paper, we explores the expressivity and trainability of the Fourier Neural Operator
(FNO). We establish a mean-field theory for the FNO, analyzing the behavior of the random …