4 results about "Parallel routing" patented technology

Disclosed is a parallel routing method for two-stage partitions using genetic algorithms, which includes: partitioning and grouping elements constituting a circuit while maintaining a connectivity between the elements; calculating a sum of weights which a connection between any two partitioned groups has; deriving a grouping in which the calculated sum of the weights has a smallest value; generating boundary boxes for respective nets constituted by elements of the derived grouping; identifying a net having a dependency with respect to the respective nets; determining a work amount required for routing each identified net; generating a partition line; partitioning two partitioned partitions into a first partition and a second partition, and placing the other partition on any one partition; and routing the first partition and the second partition in parallel by different threads.

This invention discloses a heterogeneous hybrid expert large model architecture and alignment method for high-concurrency online medical dialogue. First, the method proposes a dual-pool heterogeneous routing strategy of "4 general experts + 2 departmental experts." Through physically isolated parallel routing channels, it forces the model to simultaneously activate general language capabilities and specialized medical reasoning when processing input, fundamentally ensuring the focus and professionalism of responses. Second, it adopts a structural evolution strategy based on parameter expansion and selective freezing. By completely freezing the pre-trained basic parameters and training only newly added departmental experts, it completely eliminates catastrophic forgetting. Finally, in the model alignment stage, it employs the Group Relative Policy Optimization (GRPO) algorithm for valueless models to reduce memory overhead and designs a dual-track hybrid reward function combining low-level semantic similarity rewards and high-level structured large model referee scoring, effectively solving the "reward hacking" problem and guiding the model to generate refined responses that combine clinical accuracy and humanized interaction. This invention significantly improves the performance of medical dialogue systems. Its sparse activation and parameter freezing characteristics further ensure low latency and high throughput in online high-concurrency scenarios, providing a complete solution for the deployment of reliable medical dialogue systems.

A kind of whole network back calculation method and system of transmission network FgOTN service belong to optical communication transmission network technical field, including modeling to FgOTN service and recording service data;Service data is grouped according to time slot characteristics;Each group of service data is verified in parallel with all ports as source port;For the path of successful verification, create end-to-end FgOTN service and update service layer service data;Repeat the above steps until the whole network FgOTN service back calculation is completed.The present application avoids a large number of invalid calculations caused by point-by-point verification in traditional method through time slot characteristics grouping and batch parallel routing verification, significantly improves the FgOTN service back calculation efficiency, reduces service cut-over time, improves network operation efficiency and user experience.

This application discloses a method and system for end-to-end order-preserving scaling and scaling of stream processing based on fixed logical partitions. This method introduces a fixed logical partition layer between business keys and physical resources, establishing a two-level mapping architecture of "business key → logical partition → physical partition," thus decoupling business identity from physical location. Logical partition identifiers are encoded into message key-value pairs and transmitted transparently throughout the entire chain, enabling each layer to work collaboratively based on a unified logical view. During scaling, a three-phase consistency switching protocol is executed, and end-to-end data order preservation is ensured through dual-version parallel routing, cross-partition consumption merging, and a dual-condition confirmation mechanism. This invention resolves the core contradiction between "order preservation" and "elastic scaling" in stream processing systems, achieving strict order preservation, smooth switching, and automatic disaster recovery during dynamic adjustment of physical resources. It possesses good scalability and versatility, and is applicable to various distributed stream processing scenarios requiring strict order preservation.