Neural Network Region Division for Data Safety
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network division methods often result in data loss due to single paths leading to non-functional regions or collateral regions, causing inconsistencies in input and output operations across multiple regions.
Innovation Solution
A method and apparatus for dividing neural networks into regions by scanning operator groups, identifying broken groups, analyzing cross-region operators, and rearranging regions to ensure consistent data transmission and improved data safety through equal processor control and load balancing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If neural network models are divided into multiple collateral regions for parallel processing, then processing speed and productivity are improved, but data loss occurs when single paths lead to non-functional or earlier regions
Solution Approach 1:
The patent applies preliminary action by scanning operator groups and identifying broken groups before executing the divided neural network operations. This pre-processing step detects potential data loss paths and establishes correction mechanisms in advance, preventing data loss before it occurs during parallel processing across multiple regions
Solution Approach 2:
The patent introduces an intermediary mechanism through the identification and handling of broken operator groups. This intermediary layer acts as a mediator between the divided regions, detecting and correcting data transmission issues that occur when paths lead to non-functional or earlier regions, thus ensuring data safety while maintaining parallel processing benefits
2Productivity
If neural network models are divided into multiple regions, then parallel processing capability is enhanced, but input and output operations become inconsistent across regions
Solution Approach 1:
The patent implements feedback by scanning operator groups and identifying broken groups to detect inconsistencies in input-output operations across divided regions. This feedback mechanism provides information about data transmission issues, enabling corrective actions to be taken to restore consistency while maintaining parallel processing capabilities
Solution Approach 2:
The broken operator group identification serves as an intermediary mechanism that mediates between divided regions with inconsistent input-output operations. This intermediary detects and signals inconsistencies, enabling coordination and synchronization across regions to maintain operational consistency during parallel processing
Data Source
AI summary
A method, an apparatus, and a storage medium for dividing a neural network into regions for preventing data duplication and data loss in parallel movements of data between nodes. The method includes obtaining a neural network model comprising n operators; scanning all operator groups in the neural network model; dividing the neural network model into m regions; rescanning all operator groups in the neural network model and identifying broken operator group(s); analyzing input and output of each of the n number of operators in the broken operator group(s) and identifies operators of a specific sort; and adjusting the operators of the specific type to rearranged to keep individual inputs and outputs within a single region.