Machine Learning Model Selection for Performance and Energy Trade-Offs

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing machine learning models for tasks like speech-to-text conversion are resource-intensive, leading to high computational and energy consumption, without ensuring optimal performance thresholds are met.

Innovation Solution

A method for selecting a machine learning model that meets minimum performance thresholds while minimizing resource consumption by evaluating performance indicators and resource consumption indicators, allowing for energy-efficient and cost-effective task execution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If existing machine learning models are used for computational tasks like speech-to-text conversion, then task completion is achieved, but resource consumption and energy demand become excessively high

Engineering Contradiction:
Improvetask completionVSAvoidenergy consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary evaluation of multiple machine learning models before deployment, assessing their performance indicators and resource consumption characteristics. This advance analysis allows selection of models that meet performance thresholds while minimizing energy consumption, rather than deploying models and discovering high resource usage later.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system changes key parameters including performance thresholds, resource consumption limits, and model complexity levels to identify optimal model configurations. By adjusting these parameters and evaluating multiple models against them, the system finds models that satisfy performance requirements with reduced energy consumption compared to existing models.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If complex machine learning models are deployed to ensure high performance, then accuracy is improved, but computational resource consumption increases

Engineering Contradiction:
ImproveaccuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system applies partial action by selecting models that provide sufficient accuracy to meet defined thresholds rather than maximizing accuracy indefinitely. This approach avoids the excessive computational complexity that would result from deploying overly sophisticated models when moderate accuracy levels are adequate for the task.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system evaluates multiple machine learning models with varying complexity levels and selects those that provide adequate performance at lower computational cost. Rather than deploying a single complex high-accuracy model, the system chooses from multiple options including simpler models that meet the accuracy requirements with reduced complexity.

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

3Productivity

If multiple machine learning models are evaluated and selected based on performance and resource consumption, then optimal model selection is achieved, but selection process time increases

Engineering Contradiction:
Improvemodel selection optimizationVSAvoidselection process time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary evaluation and pre-assessment of machine learning models before final deployment decisions. By conducting advance analysis of model performance indicators and resource consumption characteristics, the system prepares model recommendations in advance, reducing the time required for final selection and deployment.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4639431B1Computer-implemented method for performing a computational task using a machine learning model
Publication Date: 2026.03.11 ELISA OYJ
  • EP4639431B1 patent drawingFigure 1
  • EP4639431B1 patent drawingFigure 2~3
  • EP4639431B1 patent drawingFigure 4~6

AI summary

According to an embodiment, a computer-implemented method (100) for performing a computational task using a machine learning model comprises obtaining (101) an indication about a computational task to be performed; obtaining (102) at least one minimum performance threshold for the computational task; obtaining (103) a plurality of machine learning models, wherein each machine learning model in the plurality of machine learning models is associated with at least one performance indicator and at least one resource consumption indicator; choosing (104) a machine learning model out of the plurality of machine learning models based at least on the at least one minimum performance threshold for the computational task, the at least one performance indicator of each machine learning model in the plurality of machine learning models and the at least one resource consumption indicator of each machine learning model in the plurality of machine learning models; and performing (105) the computational task using the chosen machine learning model.