MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641105301 A) filed by Amrita Vishwa Vidyapeetham on September 01, 2026, for A System For Chaotic Transformation Of Feature Vectors For Image Classification.
Inventors include Nair, Pramod Prabhakaran; Manusree, Anusree; and Henry, Akhila.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The present disclosure provides a system (100) for chaotic transformation of feature vectors for image classification. A convolutional network trained on few samples per class overfits and yields poorly separable features. The system (100) includes a feature extraction unit (106) extracting a feature vector, a normalization unit (108) bounding each element within an interval matched to an operating domain of a chaotic map, a chaotic transformation unit (110) applying the chaotic map element-wise to generate a transformed vector of equal dimension, and a classification unit (112) generating class probabilities through a fully connected layer (112-2). Unlike conventional approaches that enlarge a dataset or restrict model capacity, the transformation adds no trainable parameter and reshapes the feature space directly, suiting medical imaging, industrial inspection, and remote sensing where labeled samples remain scarce.
Disclaimer: Curated by HT Syndication.
