kriging
PulseAugur coverage of kriging — every cluster mentioning kriging across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New Post-stratified Importance Sampling Method Enhances Uncertainty Quantification
Researchers have introduced Post-stratified Importance Sampling (PIS), a novel method designed to address the high variance and unreliable uncertainty quantification issues inherent in traditional importance sampling te…
-
New Kriging Method Integrates Global PDE Residuals for Enhanced Probabilistic Solving
Researchers have introduced a novel method called PDE-residual-informed Kriging (PRIK) to enhance the probabilistic framework for solving partial differential equations (PDEs) using Gaussian processes. PRIK directly int…
-
Graph-informed Kriging boosts Bayesian optimization for wind farm layouts
Researchers have developed a novel Graph-informed Kriging framework to enhance the efficiency of Bayesian optimization for wind farm layouts. This new method embeds reusable aerodynamic interaction information into the …
-
New research explores training-free LLM preference learning methods · 2 sources tracked
Two new research papers explore methods for improving how large language models (LLMs) handle sequential preference learning without requiring extensive fine-tuning. The first paper, "AdaptFuse," introduces a training-f…
-
New research reveals spectral structures within Prior-Data Fitted Networks
Two new research papers explore the internal mechanisms of Prior-Data Fitted Networks (PFNs), a type of tabular foundation model. The first paper provides mechanistic evidence that PFNs capture spectral structures relat…
-
New bifidelity surrogate model uses active learning for random fields
Researchers have developed a new bifidelity Karhunen-Loève expansion (KLE) surrogate model designed for field-valued quantities of interest under uncertain inputs. This model combines the efficiency of KLE with polynomi…
-
Reinforcement learning optimizes nuclear physics experiments
Researchers have developed a novel data-driven control framework that leverages reinforcement learning and surrogate modeling to optimize target polarization in nuclear physics scattering experiments. This system uses o…
-
Transformers can approximate Gaussian Kernel Regression, research shows
A new research paper demonstrates that standard softmax-attention transformers can approximate Gaussian kernel ridge regression (KRR) predictors during their forward pass. The study constructs a single-head transformer …
-
New Gaussian process framework offers scalable, exact inference
Researchers have developed an adaptive multi-resolution Gaussian process framework designed to improve the scalability and fidelity of probabilistic machine learning models. This new approach constructs a naturally data…
-
New GP-Refiner framework accelerates Diffusion Transformers
Researchers have developed a new framework called GP-Refiner to accelerate Diffusion Transformers, a dominant paradigm in generative AI. This plug-and-play method uses Gaussian Process Regression to dynamically correct …
-
Researchers advise against using Gaussian kernels in machine learning
A recent paper argues against the widespread use of the Gaussian kernel in machine learning tasks like regression and classification. The authors contend that this kernel, also known as the squared exponential or radial…
-
New Bryson-Frazier smoother enhances Gaussian Process regression stability
Researchers have developed a modified Bryson-Frazier (MBF) smoother for temporal Gaussian Process regression. This method offers a more numerically stable alternative to the Rauch-Tung-Striebel (RTS) smoother by avoidin…
-
New generative GPR model robustly handles outliers
Researchers have developed a new generative Gaussian Process Regression (GPR) model designed to overcome the significant distortion caused by outliers in traditional GPR methods. This novel approach models observation-s…
-
New SpotOptim Python package released for black-box function optimization
The SpotOptim Python package has been released, offering a framework for optimizing expensive black-box functions. It utilizes a Kriging-based approach with Expected Improvement and supports various variable types, nois…
-
New factor graph method boosts Gaussian process regression scalability
Researchers have developed a novel factor graph approach to address the scalability challenges in multi-output Gaussian process regression. This new method expresses the regression problem as a factor graph, enabling ef…
-
New framework formalizes structure encoding in AI representations
Researchers have introduced a new framework called Legendre dynamics, which formalizes how internal representations in learning systems can encode underlying physical or statistical structure. This approach uses Legendr…
-
New Adaptive Nyström Method Enhances Gaussian Process Regression Scalability
Researchers have developed an adaptive Nyström method to improve the scalability of Gaussian Process Regression (GPR). This new approach greedily selects landmark points to minimize approximation errors, outperforming r…
-
New Bayesian Online Learning Framework Aggregates Experts for Adaptive Prediction
Researchers have developed a new framework for Bayesian online learning that addresses the challenge of fixed inferential choices by treating update rules as experts. This aggregation method competes with the best exper…
-
Gaussian Process Regression Enhances Monte Carlo Tree Search for Continuous Actions
Researchers have developed a new method for Monte Carlo Tree Search (MCTS) that utilizes Gaussian Process Regression to improve performance in environments with continuous action spaces. This approach aims to better agg…
-
New Gaussian Process method solves complex wave problems with uncertainty quantification
Researchers have developed a novel method for solving complex wave propagation problems governed by the Helmholtz equation, particularly in dissipative media where the squared wavenumber is complex. This new approach ex…