Compression Model for Circuit Parameter Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for searching global solutions in circuit design and other analysis targets face challenges due to the multimodal nature of the solution space, leading to inefficient iteration counts and potential trapping in local solutions.
Innovation Solution
A data processing apparatus utilizing a compression model learned from variable combinations to compress data dimensions, allowing for efficient acquisition of target points and change amounts within a compressed space, facilitating the optimization of circuit parameters through a regression model and iterative simulation adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simple gradient method is used for searching, then the searching process is simple, but the searching result is easily trapped in local solution instead of global solution
Solution Approach 1:
The patent transforms the original high-dimensional design space into a compressed low-dimensional space using a compression model (autoencoder). This dimensionality reduction changes the search landscape, allowing the simple gradient method to effectively navigate the compressed space and find global solutions that would be inaccessible in the original space, thereby resolving the contradiction between search simplicity and solution quality.
2Reliability
If metaheuristics such as simulated annealing or genetic algorithm are used to find global solution, then global solution can be found, but the iteration count increases and searching becomes time-consuming
Solution Approach 1:
By compressing the design space into a lower-dimensional representation, the patent reduces the complexity of the search landscape. This allows gradient-based methods to converge faster and require fewer iterations compared to metaheuristics, significantly reducing searching time while still achieving global solutions.
Solution Approach 2:
The patent replaces complex metaheuristic search mechanisms (simulated annealing, genetic algorithms) with a simpler gradient-based optimization method in the compressed space. This substitution maintains solution quality while dramatically reducing the computational overhead and iteration count associated with metaheuristics.
3Loss of information
If dimensions of data are not compressed, then all information is preserved, but the searching processing becomes inefficient in multimodal solution space
Solution Approach 1:
The patent uses a compression model to transform high-dimensional data into a lower-dimensional compressed space that preserves the essential structure and relationships of the original data. This compressed representation maintains sufficient information for effective optimization while enabling more efficient searching in the reduced-dimensional space.
Solution Approach 2:
The compression model extracts and retains only the most essential features and relationships from the high-dimensional data, discarding redundant information. This extraction process creates a compact representation that preserves critical information needed for finding global solutions while reducing the complexity of the search space.
Data Source
AI summary
A non-transitory computer-readable storage medium storing a program that causes a computer to execute a process, the process includes acquiring, based on a compression model that is acquired by learning processing on a set of data generated by using a combination of values of variables and that compresses dimensions of data, a point corresponding to data generated by using a predetermined combination of the values of variables within a compressed space; acquiring, based on the point corresponding to the data generated by using the predetermined combination, a target point within the space corresponding to a target value of a characteristic changing in accordance with the values of variables, and a regression model within the space for a predetermined variable of variables, a change amount of the predetermined variable; and changing the value of the predetermined variable included in the predetermined combination by using the change amount.