arXiv cs.LG
PulseAugur coverage of arXiv cs.LG — every cluster mentioning arXiv cs.LG across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New framework for subgraph filter learning addresses incomplete graph data
Researchers have introduced a novel framework called Subgraph Filter Learning (SFL) to address challenges in graph signal processing where complete graph topology is often unavailable. This framework proposes using subg…
-
New framework assesses query identifiability in multi-view pretraining
Researchers have developed a formal framework to determine query identifiability in multi-view pretraining scenarios, where data from multiple sources is integrated through a shared interface. They proved that ambiguity…
-
New algorithm tackles online learning of complex linear dynamical systems
Researchers have developed a new unified algorithm for online learning of Linear Dynamical Systems (LDS). This algorithm aims to achieve sublinear regret while adapting its memory usage to the intrinsic complexity of th…
-
AI models identify radioisotopes using computer vision techniques
Researchers have developed a novel machine learning approach for identifying radioisotopes in urban environments, converting gamma-ray data into spectrograms for analysis by computer vision architectures. This method en…
-
New research uses economic modeling to optimize Airbnb's marketplace dynamics
This paper explores how to understand and optimize two-sided marketplaces, using Airbnb as a primary example. It combines economic modeling and causal inference to analyze guest booking behavior based on host-set prices…
-
Geospatial AI models show poor transferability in agriculture, study finds
A new paper evaluates three geospatial foundation models—Prithvi, SpectralGPT, and SatMAE—for their effectiveness in agriculture applications. The study found that these models exhibit significant degradation when trans…
-
New research paper details curriculum learning for complex reasoning
A new research paper, "Learning to Reason with Curriculum II: Compositional Generalization," explores how breaking down complex problems into simpler sub-problems can lead to more efficient learning. The study focuses o…
-
Withdrawn arXiv paper defends cosine similarity on normalized embeddings
This paper, originally submitted to arXiv, has been withdrawn by its author. It aimed to defend cosine similarity in learned embeddings, arguing that a previously identified ambiguity related to a "gauge" matrix is only…
-
New framework unifies generative models via self-consistent paths
Researchers have introduced a new framework for understanding generative models, focusing on the concept of "self-consistent generative paths." This framework defines a path as self-consistent if it represents a random …
-
New AI architecture enables certified zero-shot inference for physical systems
Researchers have developed a novel asymmetric two-pathway architecture for physics-informed learning that aims to improve transferability across different system configurations. This approach uses a teacher encoder with…
-
LLM selective classification improved with pairwise queries
Researchers have developed a new method for selective classification in binary tasks, particularly relevant for large language models (LLMs) performing in-context learning. The technique involves using additional pairwi…
-
New framework decodes monkey whole-body movement from neural signals
Researchers have developed a new framework for recording and modeling neural activity associated with natural whole-body movements in monkeys. This system combines large-scale cortical signals with multi-view motion cap…
-
New Algorithm Enhances Bayesian Network Classifiers for Clinical Data
Researchers have developed a parallelized version of the Baymex algorithm to improve the scalability of learning discretized Bayesian Network classifiers. This enhanced algorithm adaptively steers optimization to reduce…
-
New adversarial training defends LLMs against evolving jailbreaks
Researchers have developed a novel bi-level adversarial training framework designed to defend large language models against evolving jailbreak prompts. This method simulates diverse jailbroken activations by extrapolati…
-
New bandit algorithm tackles channel noise in spectrum access
Researchers have developed a new algorithm for channel allocation in opportunistic spectrum access systems, addressing limitations of existing methods. The algorithm, a contextual multi-play multi-armed bandit, accounts…
-
New research details feature learning in wide neural networks
A new research paper explores feature learning in wide two-layer neural networks under the Maximal Update Parametrization ($\mu$P). The study establishes four key structural results, including the global existence and u…
-
Min-Max Optimization Needs Exponential Queries, Study Finds
A new research paper explores the computational complexity of min-max optimization for non-convex and non-concave functions. The study demonstrates that finding an approximate stationary point for such functions require…
-
New method decodes imagined speech from brain activity
Researchers have developed a novel method for decoding imagined speech from brain activity, addressing the scarcity of imagined speech datasets. Their approach uses paired MEG recordings from trained musicians, mapping …