#unsupervised learning

try —

15 papers match

cs.LG2026

APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems

Shentong Mo, Yatao Bian

The paper introduces Atomic Policy Optimization (APO), an unsupervised method that learns to predict 3D structures of atomic systems by optimizing a policy with dual rewards for st…

#unsupervised learning#3d structure prediction#atomic systems#policy optimization
stat.ML2026

Early Failure Prediction from Near-Anomaly Detection: A Proactive Approach

Léa Billet, L{é}a Billet, Louise Travé-Massuyès +3

The paper introduces CANARI, an unsupervised method based on the Christoffel function to identify near-anomalies—samples close to the distribution boundary—to predict failures befo…

#anomaly detection#early failure prediction#christoffel function#unsupervised learning
cs.CE2026

CANN-EUCLID: unsupervised constitutive artificial neural network model discovery from full-field data

Benjamin Alheit, Siddhant Kumar, Mathias Peirlinck

The paper introduces CANN‑EUCLID, a method that combines constitutive artificial neural networks with an unsupervised full‑field discovery framework to infer sparse hyperelastic ma…

#constitutive modeling#neural networks#unsupervised learning#full-field data
cs.LG2026

Clustering algorithms for multivariate wind farm SCADA data filtering

Nicolò Italiano, Vasilis Pettas, Tuhfe Göçmen +1

The paper evaluates several clustering algorithms for automatically filtering SCADA data from wind turbines to separate normal operation from anomalies, introducing robust evaluati…

#wind farm data#scada filtering#clustering algorithms#anomaly detection
cs.NE2026

Visual Place Recognition Using Rate-Encoded Spiking Neural Networks with Discrete STDP Learning

Altzi Tsanko, Oikonomou Katerina Maria, Antonios Gasteratos

The paper presents a tensor‑native implementation of a spiking neural network for visual place recognition that uses discrete STDP learning, and shows how design choices like deter…

#visual place recognition#spiking neural networks#unsupervised learning#STDP
cs.LG2026

FastCentNN: Accelerating Centroid Neural Network with Entropy Proxy

Le-Anh Tran

The paper introduces FastCentNN, an accelerated version of the Centroid Neural Network that uses an early splitting strategy based on a training entropy proxy to reduce unnecessary…

#unsupervised learning#clustering#online learning#algorithm acceleration
cs.LG2026

An Efficient Newton Algorithm for Nonnegative Matrix Factorization with the Kullback-Leibler Divergence

Damien Lesens, Jérémy E. Cohen, Bora Uçar

The paper proposes a Newton-type algorithm for nonnegative matrix factorization using the Kullback-Leibler divergence, employing a second‑order Taylor expansion and a generalized H…

#nonnegative matrix factorization#kullback-leibler divergence#newton method#halS algorithm
stat.ME2026

A Leave-One-Out Influence Statistic for Density-Based Outlier Detection

Aurélien Nicosia, Thierry Duchesne, Michel Carbon

The paper introduces a computationally efficient leave-one-out influence score for density-based unsupervised outlier detection using the Linear-Blend Frequency Polygon estimator,…

#outlier detection#density estimation#leave-one-out#unsupervised learning
cs.CV2026

Statistical Non-linear Reconstruction Loss for Image Anomaly Detection

Nguyen Minh Tri, Hoang Khuong Duy, Huynh Cong Viet Ngu

The paper introduces a non-linear reconstruction loss with a sigmoid squashing function and a statistical calibration method to reduce outlier leakage in unsupervised image anomaly…

#image anomaly detection#unsupervised learning#reconstruction loss#industrial inspection
cs.RO2026

UR-VC: Unsupervised Robotic Value Correction for Time-Derived Progress Proxies

Lirui Zhao, Modi Shi, Li Chen +3

The paper proposes UR-VC, an unsupervised method that improves noisy time‑derived progress labels in robot demonstrations by aggregating timestamps from similar states across episo…

#unsupervised learning#progress estimation#manipulation#demonstration data
astro-ph.HE2026

Semi-supervised morphological classification of fast radio bursts from the second CHIME/FRB catalogue

Bo Lin Fan, Renée Hložek, Antonio Herrera Martin

The paper introduces a convolutional autoencoder–based unsupervised classifier to group fast radio bursts from the CHIME/FRB catalogue into morphological classes and to predict rep…

#fast radio bursts#morphological classification#unsupervised learning#autoencoders
cs.CV2026

A Masked Autoencoder Approach to Unsupervised Steel Surface Defect Recognition

Shrey Patel

The paper introduces a transformer‑based masked autoencoder that learns visual representations from unlabeled steel surface images and uses clustering to recognize defect types wit…

#unsupervised learning#masked autoencoders#steel surface inspection#representation learning
cs.SD2026

Trajectory Variance: An Unsupervised Measure of Developmental Vocal Plasticity in Birdsong

Kanghwi Lee

The paper introduces trajectory variance, an unsupervised metric that quantifies how much a bird vocalization would change across developmental ages by using a displacement model t…

#unsupervised learning#birdsong analysis#developmental vocal plasticity#autoencoders
cs.RO2026

Enabling 24-hour Agricultural Robotics: Unsupervised Day-to-Night Cross-Modal Image Translation for Nighttime Visual Navigation

Robel Mamo, Rajitha de Silva, Grzegorz Cielniak +1

The paper presents an unsupervised image translation method that converts daytime RGB images of crop rows into nighttime near‑infrared images, allowing daytime semantic labels to b…

#day-to-night image translation#unsupervised learning#semantic segmentation#nighttime visual navigation
cs.AI2026

Connected by Construction: Learning Tractable Near-Tour Marginals for Traveling Salesman Problems

Ke Sun, Xinyuan Zhang, Xinwu Qian

The paper introduces C2TSP, an unsupervised learning framework that directly models near‑tour edge marginals for the traveling salesman problem using a connected‑by‑construction 1‑…

#traveling salesman problem#combinatorial optimization#unsupervised learning#graphical models

One search, two signals: results blend meaning (embedding similarity, so papers that never use your words still surface) with keyword matches on titles, abstracts and summaries. Free, no sign-in needed.