18056379. DISTRIBUTED MACHINE LEARNING COMPILER OPTIMIZATION simplified abstract (QUALCOMM Incorporated)

DISTRIBUTED MACHINE LEARNING COMPILER OPTIMIZATION

Organization Name

QUALCOMM Incorporated

Inventor(s)

Weiliang Zeng of San Diego CA (US)

Christopher G. Lott of San Diego CA (US)

Edward H. Teague of San Diego CA (US)

Yang Yang of San Diego CA (US)

Joseph Binamira Soriaga of San Diego CA (US)

DISTRIBUTED MACHINE LEARNING COMPILER OPTIMIZATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 18056379 titled 'DISTRIBUTED MACHINE LEARNING COMPILER OPTIMIZATION

Simplified Explanation

The abstract describes a method for optimizing the compilation of a machine learning model to be executed on target edge devices.

• Compute nodes are allocated to a compiler optimization process for the machine learning model.
• The machine learning model has a compute graph representation with nodes as kernel operators and edges defining precedence constraints.
• Optimization rounds are scheduled for the process among the allocated compute nodes.
• A sequencing and scheduling solution is applied at each node per round to obtain a performance metric.
• The best performance metric solution is identified and implemented for execution on the target edge devices.

Potential Applications

This technology can be applied in various fields such as:

• Internet of Things (IoT) devices
• Autonomous vehicles
• Robotics

Problems Solved

This technology helps in:

• Optimizing the compilation of machine learning models for edge devices
• Improving the performance of machine learning models on target devices

Benefits

The benefits of this technology include:

• Enhanced efficiency in executing machine learning models on edge devices
• Improved performance metrics for machine learning tasks

Potential Commercial Applications

Potential commercial applications of this technology include:

• Edge computing companies
• AI hardware manufacturers
• IoT solution providers

Possible Prior Art

One possible prior art for this technology could be the use of distributed computing systems for optimizing machine learning models.

What are the specific performance metrics used to evaluate the optimization process in this method?

The specific performance metrics used to evaluate the optimization process in this method are not explicitly mentioned in the abstract. However, it can be assumed that metrics such as execution time, resource utilization, and energy efficiency could be considered.

How does this method compare to existing techniques for optimizing machine learning models for edge devices?

The abstract does not provide a direct comparison of this method to existing techniques. Further research and analysis would be needed to determine how this method differs from and improves upon existing techniques for optimizing machine learning models for edge devices.

Original Abstract Submitted

A method for optimizing the compilation of a machine learning model to be executed on target edge devices is provided. Compute nodes of a plurality of compute nodes are allocated to a compiler optimization process for a compiler of said machine learning model. The machine learning model has a compute graph representation having nodes that are kernel operators necessary to execute the machine learning model and edges that connect said kernel operators to define precedence constraints. A round of optimization is scheduled for the process amongst the allocated compute nodes. At each allocated compute node a sequencing and scheduling solution is applied per round to obtain a performance metric for the machine learning model. From each compute node the performance metric is received and a solution that has the best performance metric is identified and implemented for execution of the machine learning model on the target edge devices.