clustering

This paper compares clustering algorithms for filtering wind farm SCADA data to retain only normal operation measurements, introducing evaluation metrics for unlabeled data and providing recommendations based on tests on three offshore turbines.

This paper proposes a learnable Dirichlet-process cache that allocates memory slots only for novel inputs, enabling efficient associative recall with a cache size proportional to distinct items rather than token count. It combines DP-means clustering with recurrent backbones, demonstrating effectiveness on associative recall benchmarks and real-world streams.

This paper presents a nonparametric Bayesian inverse reinforcement learning approach using a Dirichlet process prior to infer multiple latent reward types from expert demonstrations, implementing a collapsed Gibbs sampler with parallelization via Ray for scalability.

Proposes CAGI, a framework that integrates clustering and generative adversarial networks to improve missing data imputation by exploiting latent subgroup structures, achieving superior performance on benchmark datasets.

This paper presents an unsupervised clustering-based framework using K-Means++ to detect suspicious trading patterns in capital market data, achieving a silhouette score of 0.561 and identifying 2.02% of trades as potentially fraudulent.

Introduces PRISM, a prototype language model architecture that uses sparse, non-negative mixtures of learned prototypes for interpretable sequence modeling, achieving competitive performance and enabling fast training data attribution and model editing without fine-tuning.

Proposes scKDGM, a framework that uses KAN-guided dynamic graph masked learning and cross-view contrastive learning for clustering single-cell RNA-seq data, achieving state-of-the-art performance on 12 real datasets.

Proposes a two-stage cascaded framework for cost-aware LLM serving that clusters queries and routes them to cost-effective models, then escalates low-quality outputs to stronger models. Retains 97-99% of accuracy while reducing inference cost.

Proposes a term-centric framework for inducing hierarchical taxonomies from heterogeneous text sources, enabling cross-source alignment and interpretable hierarchies. Experiments on a multi-source benchmark demonstrate improved coherence and quality over text- and summary-based baselines.

This paper introduces a method called 'Learn to Cluster' to quantify and interpret social interactions among pedestrians for better trajectory prediction. It uses probabilistic latent variable generative learning to cluster social interactions without labels, improving robustness for autonomous driving and social robots.

Flash-KMeans is an IO-aware implementation of exact KMeans that redesigns the algorithm around modern GPU bottlenecks, achieving 33x speedup over cuML and 200x over FAISS by eliminating redundant memory reads and writes.

This paper proposes CRUMB, a three-stage inference wrapper that clusters test queries and selects a distributionally matched training subset via MMD minimization to enable efficient Prior-Fitted Network inference on large datasets, achieving state-of-the-art context selection on 51 TabArena datasets.

Yagil Bubrovnik presented at WWDC, demoing LM Studio's upcoming clustering feature on stage, crediting the MLX team for their work.

Sigma-Branch restructures pretrained dense networks into a hierarchical binary tree with a shared backbone, routers, and specialized leaves, reducing per-inference active parameters by 58–60% while staying within 1.72 pp of baseline accuracy on CIFAR-100, ImageNet-1K, and ModelNet40.

FailureScope is a behavioral-diagnosis method that clusters evaluation probes by cross-model pass/fail patterns to create interpretable failure taxonomies, validated across single-turn benchmarks, multi-turn dialogue, and adversarial agent attacks.

Proposes SSD-FL, a serverless semi-decentralized federated learning methodology that optimizes cluster formation in heterogeneous environments using effective loss functions and Cheeger inequality-based iterative clustering, improving convergence and communication efficiency.

The author announces a new blog post on clustering three Jetson Nano Orin Supers for distributed training and inference, continuing a series to help people build small compute clusters with accessible hardware.

Braintrust's Topics feature uses LLM summarization to make production agent traces tractable for clustering and classification at scale, inspired by Anthropic's Clio approach.

ZipSplat is a token-based feed-forward 3D Gaussian Splatting model that uses k-means clustering to decouple Gaussian placement from the pixel grid, achieving ~6x fewer Gaussians while setting new state-of-the-art results on DL3DV and RealEstate10K without requiring ground-truth poses or intrinsics.

PE-means adapts the private evolution algorithm to differentially private k-means clustering, achieving a 20% average improvement in clustering loss over existing methods.