Graph Structure Acquisition for Efficient Network Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning requires an inefficient process to search for high-performance network structures, leading to repetitive trial and error among researchers, as existing techniques do not adequately disclose effective network search methods.
Innovation Solution
An information processing device with a structure acquisition unit that acquires and generates graph structures based on information from search history and structure information databases, enabling efficient search for network structures through graph-structured networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If researchers perform trial and error to search for network structures, then they can find high-performance networks, but the process is inefficient and time-consuming
Solution Approach 1:
The system performs preliminary actions by automatically searching for and evaluating network structures before researchers need to do trial and error. The structure acquisition unit pre-acquires graph structures and the performance evaluation unit pre-evaluates them, so when researchers need a network structure, the work has already been done in advance.
Solution Approach 2:
The system implements feedback by evaluating network structures and using the evaluation results to guide further search. The performance evaluation unit provides feedback on which structures work well, and this information is used to refine the search process and avoid repeating unsuccessful approaches.
2Reliability
If researchers independently search for network structures, then they can find solutions, but the same trial and error is repeated across different research groups
Solution Approach 1:
The system merges the search efforts of multiple researchers by centralizing the network structure search in one system. Instead of each researcher independently searching, the structure acquisition unit acquires structures that can be shared and reused, combining resources to avoid duplicate work.
Solution Approach 2:
The system creates copies of successful network structures and makes them available for reuse. Once a high-performance structure is found, it can be copied and applied to different problems or research contexts, eliminating the need to rediscover the same structures repeatedly.
3Adaptability or versatility
If deep learning uses complex network structures, then expression capabilities improve, but the complexity of searching for optimal structures increases
Solution Approach 1:
The system performs self-service by automatically searching for and evaluating network structures without requiring manual design. The structure acquisition unit and performance evaluation unit work together to autonomously find suitable structures, reducing the complexity burden on researchers while maintaining high expression capabilities.
Data Source
AI summary
There is provided an information processing device capable of searching for a network structure more efficiently, the information processing device including: a structure acquisition unit configured to acquire a graph structure searched for on a basis of information related to a structure of a graph-structured network.