Skip to main content
GameDev.net gamedev.net

Machine_Learning News

The latest Machine_Learning coverage curated for game developers.

ComboStoc: Combinatorial Stochasticity for Diffusion Generative Models

The introduction of ComboStoc offers a significant advancement for diffusion generative models, particularly beneficial for graphics programmers. By addressing combinatorial complexity, this approach enhances training efficiency and test-time generation, allowing …

Research arXiv cs.GR · 4 months, 4 weeks ago
STEP-Parts: Geometric Partitioning of Boundary Representations for Large-Scale CAD Processing

The STEP-Parts toolchain offers a significant advancement for CAD processing by directly extracting geometric instance partitions from raw STEP B-Reps. This approach preserves analytic surface structure and topological adjacency, which …

Research arXiv cs.GR · 5 months, 1 week ago
Towards Extended Reality Intelligence for Monitoring and Predicting Patient Readmission Risks

The integration of mixed reality (MR) in healthcare could revolutionize how developers approach user interfaces. A recent study highlights the use of machine learning to predict patient readmission risks, achieving …

Research arXiv cs.GR · 6 months ago
End-to-End Training for Unified Tokenization and Latent Denoising

The introduction of UNITE, an innovative autoencoder architecture, streamlines the training of latent diffusion models by combining tokenization and generation into a single stage. This breakthrough could significantly reduce the …

Research arXiv cs.GR · 6 months ago
Masked BRep Autoencoder via Hierarchical Graph Transformer

A new self-supervised learning framework for CAD models could significantly enhance the efficiency of graphics programming and design tasks. By leveraging a masked graph autoencoder and a hierarchical graph Transformer, …

Research arXiv cs.GR · 6 months, 1 week ago
GCTAM: Global and Contextual Truncated Affinity Combined Maximization Model For Unsupervised Graph Anomaly Detection

The GCTAM model introduces a significant advancement in unsupervised graph anomaly detection, achieving a 15-20% improvement over previous methods. This is particularly relevant for developers working with large datasets, as …

Research arXiv cs.GR · 6 months, 3 weeks ago
BRepMAE: Self-Supervised Masked BRep Autoencoders for Machining Feature Recognition

BRepMAE introduces a self-supervised framework for machining feature recognition in CAD models, achieving high accuracy with minimal data. This advancement is particularly relevant for programmers and designers working with CAD …

Research arXiv cs.GR · 7 months ago