MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202641110371 A) filed by Vinoth Kumar Chandrasekaran on September 15, 2026, for A Computer-Implemented System And Method For Determining Search Breadth For Nearest- Neighbour Retrieval Under Multi-Version Concurrency Control.
Inventor includes Vinoth Kumar Chandrasekaran.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: A computer-implemented database computing system and method for executing approximate nearest-neighbour (ANN) similarity queries over proximity-graph indexes under Multi-Version Concurrency Control (MVCC) without silent retrieval shortfall. The system maintains transaction validity records (such as MVCC deletion bitmaps) for indexed points. Prior to commencing graph traversal, a vector query execution engine derives an optimal search breadth parameter (ef) guaranteeing that the probability of retrieving at least k snapshot-visible results is at least equal to a specified confidence level R under a binomial distribution evaluated with visible fraction p. The breadth is derived in closed form via the positive root of a quadratic equation in u = sqrt(ef), accounting for binomial variance. Graph traversal expands dead nodes to maintain metric-space navigability without edge repairs, emitting only visible nodes. A secondary fallback resolves local visibility collapse bounded by an index- cardinality proportional escalation cap.
Disclaimer: Curated by HT Syndication.
