cs.LG
PulseAugur coverage of cs.LG — every cluster mentioning cs.LG across labs, papers, and developer communities, ranked by signal.
20 day(s) with sentiment data
-
Researcher seeks arXiv endorsement for ML preprint on noisy sensor data
A researcher is seeking an endorsement to publish a preprint on arXiv concerning a machine learning study. The paper focuses on learning stable latent manifolds from noisy sensor data, presenting an algorithm with bound…
-
New programming language Cajal compiles discrete algorithms to recurrent neurons
Researchers have developed a new programming language called Cajal that allows discrete algorithms, such as iteration and conditionals, to be expressed in a differentiable form compatible with gradient-based learning. T…
-
LoCaLS algorithm enables efficient local causal discovery with latent variables
Researchers have introduced LoCaLS, a novel algorithm for local causal structure learning that addresses limitations in existing methods. Unlike approaches that require learning the entire global causal structure, LoCaL…
-
New Bayesian Optimization Algorithms Tailored for Probability Simplex
Researchers have developed a new family of Bayesian optimization algorithms, termed \"$\alpha$-GaBO\", specifically designed for optimizing functions over the probability simplex. This novel approach leverages informati…
-
New SHRED-ROM method enables real-time optimal control for complex systems
Researchers have developed a novel method called SHallow REcurrent Decoder networks-based Reduced Order Modeling (SHRED-ROM) to create real-time optimal controllers for complex dynamical systems. This approach leverages…
-
New ALAS kernel family enhances Bayesian optimization flexibility
Researchers have introduced ALAS, a novel family of Gaussian Process kernels designed for flexible Bayesian optimization. ALAS utilizes symmetric alpha-stable spectral components, allowing it to adapt its effective smoo…
-
New framework enhances flight diversion prediction using generative AI
Researchers have developed a novel framework to address the scarcity of flight diversion data in aviation records, which hinders the training of predictive machine learning models. The proposed solution involves a gener…
-
New PLayer-FL method improves personalized federated learning
Researchers have developed PLayer-FL, a novel approach to personalized layer-wise cross-silo federated learning designed to improve performance on non-IID data. Unlike previous methods that used ad-hoc heuristics, PLaye…
-
New research quantifies cost of per-class coverage under distribution shift
Researchers have characterized the cost of achieving valid per-class coverage in recognition systems when distribution shift occurs between training and testing data. They found that while split conformal prediction mai…
-
New unified approach yields bounds for contractive stochastic approximation
Researchers have developed a novel, unified approach to establish mean-square and concentration bounds for stochastic approximation (SA) algorithms. This method addresses contractive mappings in arbitrary norms and mult…
-
New geometric approach optimizes SVM hyperplanes iteratively
Researchers have developed a novel iterative geometric approach to optimize separating hyperplanes, specifically for the hard-margin Support Vector Machine (SVM) classifier. This method aims to improve the efficiency of…
-
New theory addresses domain generalization challenges in machine learning
A new research paper introduces Hierarchical Domain Generalization, a theoretical framework for machine learning models to extrapolate beyond finite observed data regions. The study posits that the primary challenge in …
-
New research explores causal effect identifiability and heterogeneity
Two new research papers explore advanced concepts in causal inference, focusing on the identifiability of causal effects under varying conditions. The first paper, "On the Granularity of Causal Effect Identifiability," …
-
New lower bound shows bandit convex optimization is harder than linear bandits
Researchers have established a new lower bound for bandit convex optimization, demonstrating that it is fundamentally more complex than linear bandits. The study introduces a novel class of convex functions that reveal …
-
New method enhances causal discovery for irregular time series data
Researchers have developed an extension to the PCMCI+ method to enable causal discovery on irregularly sampled time series data. This new approach aggregates causal influence over temporal windows, overcoming the limita…
-
New method optimizes energy efficiency in decentralized federated learning
Researchers have developed a new method for designing mixing matrices to improve the energy efficiency of decentralized federated learning (DFL) in wireless networks. This approach specifically targets the minimization …
-
New formulation gap identified in physics-informed learning for multiscale equations
Researchers have identified a statistical formulation gap in physics-informed learning for nonlinear multiscale elliptic equations. They proved a finite-sample error bound for a variational neural solver, showing that s…
-
New framework for testing probability distributions unveiled
Researchers have developed a new framework for testing probability distributions, particularly in high-dimensional or continuous domains. This approach, which uses samples from an unknown distribution to compare it agai…
-
New Weak Penalty NODE method improves chaotic system modeling from noisy data
Researchers have developed a new method called Weak Penalty NODE to improve the accuracy of Neural Ordinary Differential Equations (Neural ODEs) when modeling chaotic dynamical systems from noisy time series data. This …
-
Temporal graph learning models show mixed results in capturing graph characteristics
A new research paper published on arXiv investigates the learning capabilities of temporal graph learning models. The study systematically evaluates eight models across eight fundamental graph characteristics, including…