MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202621097509 A) filed by Bhupendra Kumar on August 12, 2026, for A Computer-Implemented System And Method For Minority-Aware Feature Certification, Sample- Sufficiency Control, And Pre-Training Governance In Rare-Event Machine Learning.
Inventor includes Bhupendra Kumar.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention relates to a computer-implemented system (300) for minority-aware feature certification and pre-training governance in rare-event machine learning. The system (300) computes, for each candidate feature, a tie-aware two-sample separation statistic and a direction-independent edge value, a single simultaneous uncertainty radius across all candidate features, and signal and stability lower confidence bounds, and assigns a strict certification state only where both are satisfied. The evidence is stored in a persistent versioned certified-feature registry (316), and a pre-training policy gate (318) applies the registry to control admission of requested candidate features into a machine-learning model-training pipeline (330) through allow, warn, named-approval, or block decisions without silent fallback. New labelled data is incorporated by an exact atomic incremental-update module (320), and a drift and re-certification monitor (322) triggers re-evaluation upon a detected certificate state transition.
Disclaimer: Curated by HT Syndication.
