MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641084491 A) filed by Amrita Vishwa Vidyapeetham on July 09, 2026, for Classification Of Imbalanced Datasets.
Inventors include Nair, Pramod Prabhakaran; Manusree, Anusree; and Henry, Akhila.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: The present disclosure provides a system (100) and method (200) for hybrid semi-supervised classification of imbalanced datasets. A processor (106) transforms raw data points (104-4) into signal matrices (106-2), which a chaos unit (108) converts into firing-rate vectors (108-2) through execution of a one-dimensional chaotic map (108-4-2). These firing-rate vectors (108-2) encapsulate nonlinear dynamics (108-2-4) to increase inter-class separation within a high-discrimination feature space. A classifier (110) computes confidence scores (110-2) for unlabelled instances and iteratively expands a sample pool (102-6) by isolating high-confidence samples (110-4) satisfying a prediction threshold (104-2). This cascading sequence structurally hardens decision boundaries, neutralizing confirmation bias in low-labelled contexts. Unlike conventional semi-supervised models that suffer from error propagation and high annotation costs, the present approach implements a deterministic loop-back mechanism to achieve synergistic performance gains and superior computational efficiency in restricted data regimes.
Disclaimer: Curated by HT Syndication.
