Search Results: XGBoost

A novel approach using a knowledge graph with 207,000 edges across operational, financial, social, and environmental layers doubled stock-ranking accuracy after market drops, with social connections proving surprisingly insignificant. The creator seeks community insights on graph normalization and heterogeneous edge handling as they transition from research to production.

LimiX-2M builds on the ambitious LimiX-16M to deliver a unified, compact tabular foundation model that’s 3.6× faster and 4× smaller than TabPFN-2.5—without sacrificing competitive performance. For developers and researchers drowning in bespoke AutoML stacks and model sprawl, this is a serious bid to redefine how we do tabular ML in practice.