18012292. Portion-Specific Model Compression for Optimization of Machine-Learned Models simplified abstract (GOOGLE LLC)

Portion-Specific Model Compression for Optimization of Machine-Learned Models

Organization Name

GOOGLE LLC

Inventor(s)

Yicheng Fan of Mountain View CA (US)

Jingyue Shen of Santa Clara CA (US)

Deqiang Chen of San Jose CA (US)

Peter Shaosen Young of Mountain View CA (US)

Dana Alon of Mountain View CA (US)

Erik Nathan Vee of San Mateo CA (US)

Shanmugasundaram Ravikumar of Piedmont CA (US)

Andrew Tomkins of Menlo Park CA (US)

Portion-Specific Model Compression for Optimization of Machine-Learned Models - A simplified explanation of the abstract

This abstract first appeared for US patent application 18012292 titled 'Portion-Specific Model Compression for Optimization of Machine-Learned Models

Simplified Explanation: The patent application discusses systems and methods for compressing and optimizing machine-learned models by focusing on specific portions of the model.

• Key Features and Innovation:

   - Obtaining data on compression schemes for different model portions
   - Evaluating cost functions to select compression schemes
   - Applying selected compression schemes to obtain a compressed machine-learned model

• Potential Applications:

The technology can be applied in various fields where machine learning models are used, such as healthcare, finance, and autonomous vehicles.

• Problems Solved:

This technology addresses the need for efficient compression and optimization of machine-learned models to improve performance and reduce computational resources.

• Benefits:

   - Improved model efficiency and performance
   - Reduced computational resources and storage requirements
   - Enhanced scalability and applicability of machine learning models

• Commercial Applications:

"Portion-Specific Compression and Optimization of Machine-Learned Models" can be utilized in industries such as e-commerce for personalized recommendations, cybersecurity for threat detection, and manufacturing for predictive maintenance.

• Prior Art:

Readers can explore prior research on machine learning model compression techniques, optimization methods, and applications in specific industries.

• Frequently Updated Research:

Stay informed about the latest advancements in machine learning model compression and optimization techniques through academic journals, conferences, and research papers.

Questions about Machine-Learned Model Compression and Optimization: 1. How does portion-specific compression improve the efficiency of machine-learned models? 2. What are the potential challenges in implementing compression schemes for different model portions?

Original Abstract Submitted

Systems and methods of the present disclosure are directed to portion-specific compression and optimization of machine-learned models. For example, a method for portion-specific compression and optimization of machine-learned models includes obtaining data descriptive of one or more respective sets of compression schemes for one or more model portions of a plurality of model portions of a machine-learned model. The method includes evaluating a cost function to respectively select one or more candidate compression schemes from the one or more sets of compression schemes. The method includes respectively applying the one or more candidate compression schemes to the one or more model portions to obtain a compressed machine-learned model comprising one or more compressed model portions that correspond to the one or more model portions.