18119488. MULTI-PLATFORM NEURAL NETWORK DEPLOYMENT simplified abstract (Adobe Inc.)

MULTI-PLATFORM NEURAL NETWORK DEPLOYMENT

Organization Name

Adobe Inc.

Inventor(s)

Zichuan Liu of San Jose CA (US)

Xin Lu of Saratoga CA (US)

Wentian Zhao of San Jose CA (US)

MULTI-PLATFORM NEURAL NETWORK DEPLOYMENT - A simplified explanation of the abstract

This abstract first appeared for US patent application 18119488 titled 'MULTI-PLATFORM NEURAL NETWORK DEPLOYMENT

Simplified Explanation:

The patent application describes a method for converting a machine learning model for execution by a computing device. This involves generating a computing graph based on the model, detecting sub-structures within the graph, and optimizing the graph for efficient execution.

Key Features and Innovation:

• Generation of a computing graph based on a machine learning model.
• Detection and combination of sub-graphs within the computing graph.
• Optimization of the computing graph for improved performance.
• Creation of net-list and weight objects for inferencing operations.

Potential Applications: The technology can be applied in various fields such as image recognition, natural language processing, and autonomous driving systems.

Problems Solved: The technology addresses the need for efficient execution of machine learning models on computing devices.

Benefits:

• Improved performance and efficiency of machine learning models.
• Enhanced inferencing operations on computing devices.

Commercial Applications: Potential commercial applications include AI-powered software, autonomous systems, and data analytics platforms.

Questions about the Technology: 1. How does the technology optimize the computing graph for efficient execution? 2. What are the potential limitations of converting machine learning models for execution on computing devices?

Frequently Updated Research: Stay updated on advancements in machine learning model optimization and inferencing operations for computing devices.

Original Abstract Submitted

In various examples, a machine learning model is converted for execution by a computing device. For example, a computing graph is generated based on the machine learning model and sub-graphs within the computing graph that match sub-structures that are detected and combined into a vertex to generate an optimized computing graph. A net-list object and weight object are then generated based on the optimized computing graph and provided to the computing device to enable inferencing operations.