Neural tangent kernel
PulseAugur coverage of Neural tangent kernel — every cluster mentioning Neural tangent kernel across labs, papers, and developer communities, ranked by signal.
- 2026-05-13 research_milestone Publication of a paper introducing a force-aware Neural Tangent Kernel for active learning of MLIPs. source
7 day(s) with sentiment data
-
New Transport Neural Tangent Kernel Enhances Active Learning for Generative Models
Researchers have developed a new active learning framework for conditional generative models, addressing the limitations of existing methods that focus solely on scalar or vector-valued outputs. The proposed transport N…
-
New PGD algorithm sharpens neural network regression rates
Researchers have developed a new Preconditioned Gradient Descent (PGD) algorithm for training over-parameterized neural networks in nonparametric regression tasks. This method, combined with early stopping, achieves a s…
-
Neural Tangent Kernel Analysis Sheds Light on PDF Training Dynamics
Researchers have analyzed the training dynamics of feed-forward neural networks using a Neural Tangent Kernel (NTK) approach applied to Parton Distribution Functions (PDFs) derived from deep inelastic scattering (DIS) d…
-
Neural Tangent Kernels Probe Learned Materials Representations
Researchers have developed a method using the neural tangent kernel (NTK) to understand the internal representations learned by neural networks that predict material properties. By evaluating the NTK at trained paramete…
-
New theory explains Low-Rank Adaptation in large models
A new paper explores the theoretical underpinnings of Low-Rank Adaptation (LoRA), a widely used technique for efficient fine-tuning of large pretrained models. The research proposes a novel approach that bypasses the li…
-
New theory explains adversarial training benefits for physics-informed neural networks
A new theoretical framework analyzes how adversarial training, specifically using Generative Adversarial Networks (GANs), can improve the performance of Physics-Informed Neural Networks (PINNs). This research, based on …
-
New KPSA method enhances ML model interpretability efficiency
Researchers have developed a new method called Kernelized Parameter Sensitivity Atlas (KPSA) to improve the efficiency of interpretability in machine learning models. KPSA leverages the empirical neural tangent kernel t…
-
New method Learnable Spectral Activations improves neural network representations
Researchers have introduced Learnable Spectral Activations (LSA), a novel approach to implicit neural representations (INRs). LSA replaces fixed nonlinearities with a residual truncated Fourier series, allowing harmonic…
-
New initialization strategy tackles spectral bias in physics-informed neural networks
Researchers have developed a new initialization strategy for Fourier Feature Physics-Informed Neural Networks (PINNs) to address spectral bias, a common issue where certain frequencies of a target function converge slow…
-
New machine unlearning methods tackle bias and data deletion in time-series forecasting
Two new research papers introduce novel machine unlearning techniques tailored for time-series forecasting. The first paper, SSU-LSF, addresses bias in land surface forecasting models like Mamba by developing a framewor…
-
New SBD framework challenges causality in language models
A new research paper proposes the System Behavior (SBD) framework, which incorporates system behavior as a fundamental component alongside data distribution. The framework theoretically identifies a "Causality Tax" phen…
-
New method makes NTK statistics practical for large neural networks
Researchers have developed a method to efficiently estimate statistics of the Neural Tangent Kernel (NTK) for large neural networks, overcoming the computational and memory challenges of explicit calculation. This new a…
-
New framework analyzes encoder-decoder operator learning via limiting kernels
This paper introduces a novel framework for analyzing operator learning within encoder-decoder architectures. It formulates operator learning on function spaces, addressing the challenge of finite-dimensional training d…
-
Latent-MoE architecture enhances physics-informed neural networks for complex PDEs
Researchers have developed Latent-MoE, a novel domain-aware Mixture-of-Experts architecture designed to improve the performance of physics-informed neural networks (PINNs) on partial differential equations (PDEs) with c…
-
New algorithms enhance neural field reconstruction with NTK and meta-learning
Researchers have developed three new algorithms to improve the reconstruction quality and efficiency of neural fields, which map continuous coordinates to signals like color or density. The first algorithm, NTK-KIP, use…
-
AdaptNTK framework enhances AI for molecular dynamics simulations
Researchers have developed AdaptNTK, a novel framework for quantifying uncertainty and implementing active learning in neural network potentials. This single-model approach uses a regularized Mahalanobis distance in emp…
-
Recurrent Neural Network Learning Dynamics Near Bifurcations Analyzed
A new research paper explores the dynamics of learning in recurrent neural networks (RNNs) near critical transition points, known as bifurcations. The study utilizes the global empirical Neural Tangent Kernel (GeNTK) to…
-
New theory links correlation propagation to Neural Tangent Kernel in deep learning
Researchers have established a theoretical link between correlation propagation and the Neural Tangent Kernel (NTK) in deep neural networks. By combining mean-field and random matrix theories, they demonstrated that cor…
-
New tools explain neural network training for power system dynamics
Researchers have developed new analytical tools to explain the training performance of machine learning surrogate models used in power system dynamics. By adapting small-signal eigenvalue analysis from power systems, th…
-
Weak correlations principle explains linearization in gradient-based learning systems
A new paper published on arXiv explores the principle of weak correlations as the underlying reason for the linearization observed in gradient-based learning systems. The research suggests that the simplified dynamics s…