MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202647124194 A) filed by Google LLC on October 16, 2026, for Decomposing Neural Network Models On An Edge Device.
Inventors include Kaliamoorthi, Prabhu; Garud, Hrishikesh Deepak; and Song, Chen.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: Methods and systems, including computational instructions/programs encoded on computer-readable media, are described for selecting and loading decomposable expert neural network models on an edge device to implement a generative inference process locally on the device. A system receives a request representing an input condition. For each step of the generative inference process: the system selects an expert model from multiple expert models using an identifier for the expert model. Based on the identifier, the system loads data that represents the expert model into a main memory of the integrated hardware circuit. The expert model executes a computational iteration in the step of the generative inference process based on the data, the input condition, and an input image. The system computes a generative inference output in response to executing each step of the generative inference process.
Disclaimer: Curated by HT Syndication.
