Intellectual Property India Publishes Patent Application for System And Method For Adaptive Vector Quantization Of Deep Neural Networks Using Machine Learning

MUMBAI, India, July 31 -- Intellectual Property India has published a patent application (202411100475 A) filed by Mercedes-Benz Group AG on December 18, 2024, for System And Method For Adaptive Vector Quantization Of Deep Neural Networks Using Machine Learning.

Inventors include Anamika Jha; Aratrik Chattopadhyay; Mrinal Banerji; and Disha Jain.

The application for the patent was published on July 24, 2026, under issue no. 30/2026.

Abstract: The present disclosure provides a system (106) and a method (400) for adaptive quantization of deep neural networks using machine learning. The system (106) represents an input through a first neural network layer (304-1 to 304-N). The system (106) assigns a set of weight variables (306-1 to 306-N) to the first neural network layer (304-1 to 304-N) and dynamically allocates one or more equal bit-width values to generate a quantized neural network layer (304-1 to 304-N). The system (106) computes a performance loss function 318 and a fidelity loss function 320 based on the first neural network layer (304-1 to 304-N) and the quantized neural network layer (326-1 to 326-N). The system (106) optimizes a variation between a size 314 and a quantized size 316 through the computation of a loss function 324, which is based on the performance loss function 318 and the fidelity loss function 320.

Disclaimer: Curated by HT Syndication.