Using embedding functions with a deep network
Topics: AI (Deep Learning), AIOverviews, Jeff Dean, LLMO / GEO, Semantic Search
This patent describes a deep learning system that uses embedding functions to process different types of input features. The system processes categorically distinct features through specialized embedding functions that convert them into numeric representations. These are then fed into a deep neural network with multiple non-linear layers, followed by a logistic regression classifier to make predictions.
• Patent ID: US20240211759A1
• Assignee: Google LLC
• Countries: United States
• Last publishing date: Not explicitly mentioned in provided text / Assigned: April, 2024
• Inventors: Gregory S. Corrado, Kai Chen, Jeffrey A. Dean, Gary R. Holt
• Status: Pending