Neural Network Constraint Checking Across Multiple Hardware Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network design tools struggle to accommodate processing by multiple hardware devices due to hardware-specific restrictions, leading to inefficiencies and potential execution failures.
Innovation Solution
An information processing apparatus and method that includes an acquiring unit to gather hardware device restrictions and a determining unit to assess whether a neural network meets these restrictions, providing a design tool for creating a neural network compatible with multiple hardware devices, with feedback and automatic rearrangement to ensure compliance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a neural network is designed for single hardware device execution, then design simplicity is maintained, but compatibility with multiple hardware devices deteriorates
Solution Approach 1:
The patent segments the neural network design process into distinct phases: network construction, hardware restriction acquisition, and determination. This segmentation allows the system to handle multi-hardware compatibility checks as a separate, manageable step rather than integrating complexity throughout the entire design process, thus maintaining design simplicity while achieving hardware compatibility.
Solution Approach 2:
The patent applies preliminary action by acquiring hardware device restrictions and performing determination checks before the neural network is fully deployed or executed. This advance verification ensures compatibility with multiple hardware devices without requiring redesign, resolving the contradiction between design simplicity and hardware adaptability.
2Productivity
If hardware device restrictions are not checked, then design efficiency is maintained, but execution reliability deteriorates
Solution Approach 1:
The patent implements feedback by establishing a determination step that checks whether the constructed neural network satisfies hardware device restrictions. This feedback mechanism provides verification of execution reliability without significantly impacting design efficiency, as the check is performed automatically as part of the design workflow rather than requiring manual verification or iterative redesign.
3Measurement precision
If manual verification of hardware compatibility is performed, then determination accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies self-service by enabling the system to automatically acquire hardware device restrictions and perform determination checks without requiring manual verification. The determination unit automatically evaluates whether the neural network satisfies the acquired restrictions, providing accurate compatibility assessment while eliminating the time loss associated with manual checking processes.
Data Source
AI summary
There is provided an information processing apparatus which is capable of more efficiently designing a neural network which is fit for processing by a plurality of hardware devices, and an information processing method. The information processing apparatus includes: an acquiring unit configured to acquire restrictions relating to a plurality of hardware devices; and a determining unit configured to perform determination as to whether or not a neural network satisfies the restrictions.