Model Parameter Mapping for Faster High-Resolution Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Optimizing parameters of large-scale, high-resolution numerical models is a costly and time-consuming process due to the need for multiple executions with different parameter sets, especially in domains like oceanography and climate science.
Innovation Solution
A system and method using machine-learning techniques to learn a mapping between low-resolution and high-resolution model parameters through a deep neural network (PMN) and a parameter mapping transform (PMT), allowing parameter tuning at a low resolution and converting these to high-resolution equivalents, thereby reducing computational costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parameter tuning is performed at high resolution, then simulation accuracy is improved, but computational cost and time consumption increase significantly
Solution Approach 1:
The patent performs parameter tuning at low resolution first to obtain preliminary parameter sets, then uses these as initial guesses for high-resolution tuning. This preliminary action at lower computational cost guides the subsequent high-resolution optimization, reducing the number of expensive high-resolution model executions needed.
Solution Approach 2:
The patent introduces an intermediate step of low-resolution model tuning that acts as a mediator between rough initial parameters and final high-resolution parameters. The low-resolution model serves as an intermediary system that pre-processes parameter optimization before transferring results to the high-resolution model.
2Manufacturing precision
If multiple model executions with different parameter sets are performed, then parameter optimization is improved, but computational resources are consumed excessively
Solution Approach 1:
The patent segments the parameter optimization process into two distinct phases: low-resolution parameter tuning and high-resolution parameter tuning. This segmentation allows the computationally intensive optimization to be performed partially at lower resolution, reducing overall computational resource consumption while still achieving high-resolution parameter optimization.
Solution Approach 2:
The patent changes the resolution parameter of the model during the optimization process. By tuning parameters at low resolution first and then adjusting to high resolution, the system reduces computational resource usage during the exploratory phase of parameter optimization while maintaining accuracy in the final phase.
3Manufacturing precision
If high-resolution model runs are performed repeatedly, then parameter tuning accuracy is improved, but the process becomes expensive and time-consuming
Solution Approach 1:
The system performs preliminary parameter tuning at low resolution to obtain initial parameter sets that are close to optimal values. This preliminary action reduces the search space for high-resolution tuning, thereby improving tuning efficiency while maintaining parameter tuning accuracy.
Solution Approach 2:
The patent creates a low-resolution copy of the high-resolution model for preliminary parameter tuning. This simplified copy allows rapid parameter exploration and initial optimization without the computational burden of the full high-resolution model, improving overall tuning efficiency.
Data Source
AI summary
One embodiment can provide a method and system for tuning parameters of a numerical model of a physical system. During operation, the system can obtain, using a machine-learning technique, a parameter-transform model for mapping parameters of the numerical model at a first resolution to parameters of the numerical model at a second resolution, the second resolution being higher than the first resolution. The system can perform a parameter-tuning operation on the numerical model at a first resolution to obtain a first set of tuned parameters and apply the parameter-transform model on the first set of tuned parameters to obtain a second set of tuned parameters at a second resolution. The system can then generate behavior information associated with the physical system by running the numerical model at the second resolution using the second set of tuned parameters.