AI
Identifying System Failure Origins via Graph Neural Networks
A new diagnostic framework uses graph neural networks and extreme value theory to pinpoint the source and timing of system-wide failures.
October 12, 2026, 12:13 PM·Хабр
#data-science
AI
AI VK Research Unveils Scalable Graph Neural Network Method
AI VK Research has introduced a method to scale graph neural networks, reducing memory usage by over 100 times and increasing recommendation efficiency for social networks.
CNews.ru
AI
Xiaomi Introduces TabLDM for Tabular Data Analysis
Xiaomi has launched TabLDM, an AI model for tabular data that eliminates the need for task-specific fine-tuning and offers faster processing speeds compared to existing solutions.
iXBT.com
AI
Turning Steps Into Parsecs: A New Web Project
A new web application converts daily step counts into astronomical distances, using AI-generated code and data from the HYG and Gaia DR3 star catalogs.
Хабр
AI
JEPA-Anything Framework Enhances Multi-Domain Predictive Modeling
The new JEPA-Anything framework utilizes Orthogonal Factor Projection to improve capacity allocation and performance across seven distinct scientific and dynamics domains.
MarkTechPost
AI
Hybrid Quantum Model Enhances Energy Forecasting Accuracy
Silicon Quantum Computing and Schneider Electric have achieved a 20% improvement in energy forecasting accuracy using a hybrid quantum-classical model.
ForkLog
AI
NVIDIA Introduces Kumo Tabular Foundation Models
NVIDIA has launched Kumo Tabular, a series of open-source foundation models for structured data that support commercial use and feature high inference speeds.
MarkTechPost
AI
New Consortium Aims to Standardize Antibody Data for AI Training
Apheris and Ginkgo Datapoints have launched the Antibody Developability Consortium, uniting global pharma leaders to build a standardized 10,000-antibody dataset for AI training.
Tech.eu
Hardware
Benchmarking Data Processing Libraries on Limited Hardware
A performance comparison of pandas, Polars, DuckDB, and PySpark on a low-spec machine reveals how different libraries handle memory constraints and large datasets.
Хабр
AI
The Hidden Dangers of Relying on AI for Professional Tasks
Over-reliance on AI assistants for coding and analysis can erode critical thinking skills and professional expertise, turning specialists into mere operators of AI tools.
Хабр
AI
Tether's QVAC Unveils Genesis III STEM Dataset for AI Training
Tether's QVAC division has released the Genesis III synthetic STEM dataset, enabling small language models to achieve a 99.45% valid-answer rate on MMLU benchmarks.
Crypto Briefing
AI
Nums AI Launches Causilo Tabular Foundation Model
Nums AI has introduced Causilo, a tabular foundation model that leads the TabArena benchmark among single models with an Elo rating of 1794.
MarkTechPost
AI
T-Bank Launches Perseus Framework for User Event Modeling
T-Bank has released the Perseus framework for processing heterogeneous user event sequences, supporting tasks ranging from recommendation ranking to regression.
Хабр
AI
Understanding RBF SVM: A Mathematical Perspective
An overview of the RBF SVM machine learning method, covering its mathematical foundations, the role of support vectors, and techniques for handling non-linear data classification.
Хабр
AI
Google DeepMind Unveils Massive AlphaGenome Atlas for DNA Variant Analysis
Google DeepMind has released the AlphaGenome Atlas, a 1-petabyte dataset covering 9 billion human DNA variants to accelerate genomic research.
MarkTechPost
AI
Google DeepMind Unveils WeatherNext 3 for High-Resolution Global Forecasting
Google DeepMind has launched WeatherNext 3, an AI weather forecasting model that provides hourly updates with 5 km resolution for global atmospheric data.
MarkTechPost
That is everything