Machine Learning Model Configuration for In-Vehicle Hardware Constraints

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for deciding machine learning models for in-vehicle systems struggle to efficiently meet stringent constraints on power consumption, processing speed, and recognition accuracy due to the variety of processors used in different vehicles, without adequately considering hardware performance in the learning process.

Innovation Solution

An information processing method that determines a machine learning model configuration using a processor, checks if it meets hardware performance requirements, and iteratively adjusts the configuration until the requirements are met, reducing the need for subsequent evaluations and processing loads.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If machine learning model configuration is decided without considering hardware performance requirements, then model development is simpler and faster, but the model fails to meet stringent constraints on power consumption, processing speed, and recognition accuracy

Engineering Contradiction:

Improvehardware performance requirement complianceVSAvoidmodel selection process complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary determination of hardware performance requirements before executing the machine learning model. By evaluating whether the model meets power consumption, processing speed, and recognition accuracy constraints in advance, the system avoids deploying incompatible models and ensures reliability while maintaining a manageable selection process through structured pre-checks.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If multiple evaluations and iterations are performed to find a suitable machine learning model, then hardware performance requirements are met, but processing load and energy consumption increase

Engineering Contradiction:

Improvehardware performance requirement complianceVSAvoidenergy consumption during model selection

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements a feedback mechanism where the determination result of hardware performance requirements is used to control whether the machine learning model should be executed. If the model does not meet the requirements, the system provides feedback to reject or reconfigure the model, avoiding wasted energy on unsuitable deployments and reducing overall processing load through intelligent filtering.

Inventive Principle:

Principle #23Feedback

3Reliability

If machine learning model configuration is changed to meet hardware performance requirements, then power consumption and processing speed constraints are satisfied, but the number of evaluations and processing time increase

Engineering Contradiction:

Improvehardware performance requirement complianceVSAvoidmodel configuration adjustment time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent changes parameters of the machine learning model configuration, such as network depth, width, or learning rate, to satisfy hardware performance requirements. By systematically adjusting these parameters and re-evaluating against constraints, the system finds optimal configurations that meet power consumption and processing speed targets while minimizing unnecessary evaluation cycles through targeted parameter tuning.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11270228B2Information processing method and information processing system

Publication Date: 2022.03.08 PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO LTD

AI summary

An information processing method includes: deciding a configuration of a machine learning model, using a processor; performing first determination as to whether the machine learning model in the decided configuration meets a first performance requirement on hardware performance; performing learning using the machine learning model in the configuration determined to meet the first performance requirement, performing second determination as to whether a learned model obtained by the learning meets a second performance requirement on evaluation value of output of a machine learning model, and when the learned model is determined to meet the second performance requirement, outputting information indicating that both performance requirements are met; and when it is determined that the first performance requirement is not met, changing the configuration of the machine learning model, and performing the first determination as to whether the machine learning model in the changed configuration meets the first performance requirement.