面向电力多业务场景的AI训练样本分级缓存方法
By constructing a candidate sample set and a sample attribute set, evaluating and classifying the revenue density, and implementing granular reconstruction and dual-domain allocation, the memory jitter problem of caching strategies in multi-service scenarios of power systems was solved, and resource utilization was improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING YISHUNHONG INFORMATION TECH CO LTD
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-17
AI Technical Summary
In the multi-service scenarios of power systems, existing cache eviction strategies are unable to balance data access frequency and high capacity usage, resulting in memory jitter and resource waste.
By constructing a candidate sample set and a sample attribute set, converting them into cache objects and evaluating the benefit density, and then performing granular reconstruction on the objects to be reconstructed after classification, a set of reconstructed objects is generated, and a dual-domain allocation of isolated cache domain and shared cache domain is performed to generate an iterative caching strategy.
It improves the overall resource utilization of parallel training, adapts to heterogeneous data, and reduces memory jitter and resource contention.
Smart Images
Figure CN122086799B_ABST