MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106928 A) filed by S Madhumala Vinnakota; Dr Aluri Brahmareddy; Manoj Kumar Gottimukkala; Vazralu Munnangi; Eelapanti Mythri; Venkata Seshu Kiran T; and Dr B Sridhar Babu on September 05, 2026, for Neural Network Architecture For Low Latency Edge Ai Applications.
Inventors include S Madhumala Vinnakota; Dr Aluri Brahmareddy; Manoj Kumar Gottimukkala; Vazralu Munnangi; Eelapanti Mythri; Venkata Seshu Kiran T; and Dr B Sridhar Babu.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: ABSTRACT Neural Network Architecture for Low Latency Edge AI Applications The present invention relates to a neural network architecture for performing low- latency artificial intelligence inference on edge computing devices. The architecture comprises an input processing module, lightweight neural processing blocks, an adaptive precision controller, a feature reuse module, a hardware-aware execution scheduler, an intermediate memory management module and an inference controller. The lightweight neural processing blocks employ computationally efficient feature extraction and channel transformation operations. The adaptive precision controller selectively determines numerical precision for different neural processing operations based on computational complexity, hardware capability, latency requirements and accuracy requirements. The feature reuse module stores and reuses selected intermediate feature representations to reduce redundant computation and data movement. The hardware-aware execution scheduler allocates neural processing operations among available CPU, GPU, NPU, DSP, FPGA or other AI accelerator resources. The inference controller dynamically coordinates neural-network execution according to hardware conditions and target latency. The architecture thereby provides reduced inference latency, computational complexity, memory utilization and energy consumption while maintaining a desired prediction accuracy, making the invention suitable for real-time edge AI, industrial IoT, robotics, surveillance, healthcare, transportation and smart manufacturing applications.
Disclaimer: Curated by HT Syndication.
