k-nearest neighbors algorithm
PulseAugur coverage of k-nearest neighbors algorithm — every cluster mentioning k-nearest neighbors algorithm across labs, papers, and developer communities, ranked by signal.
9 day(s) with sentiment data
-
Machine learning models show improved path loss prediction for LPWANs
Researchers have conducted a systematic analysis of machine learning models for predicting path loss in Low Power Wide Area Networks (LPWANs), specifically focusing on LoRa technology. The study employed Random Forest m…
-
New ML framework enhances 6G localization with RIS and mmWave sensing
Researchers have developed a new machine learning framework for precise localization in 6G networks utilizing reconfigurable intelligent surfaces (RIS) and millimeter-wave (mmWave) sensing. This method maps received sig…
-
Machine learning models detect user deaths on social media
A new dissertation details the development of machine learning classifiers capable of automatically detecting deceased users on social networking sites. The research utilized a new dataset compiled from Wikidata and X (…
-
New 'Local Shapley' method drastically cuts data valuation computation
Researchers have introduced "Local Shapley," a novel method for data valuation that significantly reduces computational complexity. Unlike traditional approaches that consider all possible training data combinations, Lo…
-
New scale law guides detection of distribution shifts in AI embeddings
Researchers have developed a new scale law for detecting distribution shifts in high-dimensional embeddings, which constrains moment-based statistical tests. This law, derived from Chebyshev's extremal problem, suggests…
-
New POSSE-kNN method improves binary classification accuracy
A new machine learning method called POSSE-kNN has been developed for binary classification tasks, particularly for tabular data. This ensemble technique combines bootstrap sampling, random feature subspaces, out-of-bag…
-
Open-source tool optimizes AI models for lower cost and higher efficiency
Experiential Labs has released an open-source tool called World Model Optimizer (wmo) designed to reduce the cost of serving AI models. The tool optimizes agent traces to create smaller, more efficient models, and can r…
-
Ensemble classifiers boost power line outage localization performance
Researchers have explored the use of ensemble classifiers to improve the performance of line outage localization in power systems. Their study compared various ensemble methods against single-model approaches, utilizing…
-
New AI framework identifies dark vessels using SAR images and GT estimation
Researchers have developed a novel framework to identify "dark vessels" that disable their transponders for surveillance evasion. The system combines a multi-task deep learning model to predict vessel location, type, an…
-
Single-channel sEMG shows promise for efficient hand gesture recognition
Researchers have explored the use of a single surface electromyography (sEMG) channel for hand gesture classification, aiming for more efficient and low-power systems. By extracting various time-domain and frequency-dom…
-
New ASK-NN test detects LLM hallucinations via distributional drift
Researchers have developed ASK-NN, a novel asymmetric nearest-neighbor test designed to detect distributional drifts in natural language, a common indicator of hallucinations or artificial text in LLM outputs. This meth…
-
TabPFN shows promise for multimodal classification tasks
A new research paper explores the effectiveness of TabPFN as a classification head for multimodal tasks, moving beyond its traditional use in tabular data. The study found that TabPFN significantly improves calibration …
-
LLM candidate generation boosts long-tail listings on Vrbo
Researchers have developed a novel, training-free LLM-based candidate generation system designed to improve property recommendations on vacation rental platforms like Vrbo. This system addresses the challenge of the "lo…
-
New metrics assess meaningfulness of language model routing policies
Researchers have developed a new framework for evaluating language model routing policies, focusing on behavioral differentiation and stability rather than just task accuracy. They propose adapting Hierarchic Social Ent…
-
LLMs enhance GCNs for semi-supervised image classification · arXiv paper
Researchers have developed a novel method to improve semi-supervised image classification by integrating Large Language Models (LLMs) with Graph Convolutional Networks (GCNs). The approach addresses the challenge of gra…
-
ImputeViz dashboard aids missing data analysis and imputation method comparison
Researchers have developed ImputeViz, a visual analytics dashboard designed to help researchers diagnose missing data patterns and compare different imputation methods. The tool integrates popular techniques like MICE, …
-
DINOv2 underperforms SigLIP in fine-grained classification task
A Reddit user conducting a bachelor's thesis on fine-grained car classification found that the DINOv2 Giant model performed significantly worse than SigLIP2 SO400M when used as a frozen encoder for k-NN classification. …
-
New AIT-based method outperforms BERT on text classification tasks
Researchers have developed a novel method for analyzing text structure based on Algorithmic Information Theory (AIT), utilizing the Ladderpath approach to identify nested and hierarchical repetitions within sequences. T…
-
Machine learning fundamentals: supervised, unsupervised, and ensemble techniques
This article delves into fundamental machine learning concepts, covering both supervised and unsupervised learning techniques. It explores supervised learning through function approximation, the bias-variance tradeoff, …
-
New WIDER-FAIR dataset reveals bias in face detection models
Researchers have introduced WIDER-FAIR, a new dataset designed to evaluate fairness in face detection models. Built upon the WIDER-FACE benchmark, WIDER-FAIR includes manual annotations for perceived ethnicity and sex a…