Neural Network Parameter Transformation for Circuit Design Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network circuit designs face inefficiencies in machine learning processes, requiring repeated trials for optimization and performance improvement, which hampers design efficiency.
Innovation Solution
An information processing apparatus that transforms machine learning parameters from a first-type neural network to a second-type neural network, generating production information and estimate information for improved neural network circuit design, including pruning processes and nonlinear function calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If repeated trials in design and machine learning processes are conducted to improve performance, then neural network circuit performance is improved, but design efficiency deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-transforming machine learning parameters from a first-type neural network to a second-type neural network before actual circuit implementation. The transformation processing device converts weights and biases into pruned neural network parameters in advance, allowing designers to leverage existing trained models without repeating the entire machine learning process for each circuit design iteration, thus improving design efficiency while maintaining performance
Solution Approach 2:
The patent applies copying by creating a transformed version of the neural network parameter set from an existing first-type neural network. The transformation processing device copies the learned parameters and adapts them to the second-type neural network structure, enabling reuse of previously trained knowledge across different neural network architectures without requiring retraining from scratch
2Productivity
If machine learning parameters are transformed from first-type to second-type neural network, then design efficiency is improved, but parameter transformation complexity increases
Solution Approach 1:
The patent applies the intermediary principle by introducing a transformation processing device as a mediator between the first-type and second-type neural networks. This intermediary component handles the complex parameter transformation process, including weight conversion and pruning operations, isolating the complexity from both the training and deployment stages. The transformation device serves as a specialized bridge that manages the conversion complexity internally while presenting simple interfaces to users
Data Source
AI summary
An information processing apparatus receives an input of a machine learning parameter for a first-type neural network which has machine-learned an output corresponding to a predetermined input, and transforms the received machine learning parameter to a machine learning parameter for a second-type neural network which is a different type neural network from the first-type neural network. On the basis of the transformed machine learning parameter, the information processing apparatus generates production information for producing the second-type neural network, and generates estimate information regarding at least one of a scale and performance, relating to the second-type neural network produced in accordance with the production information.