7 results about "Hierarchical routing" patented technology

The application discloses a large-scale constellation hierarchical routing planning method for concurrent tasks, and belongs to the technical field of mega constellation management and routing planning.The application is used for solving the problem of constellation routing planning under large-scale concurrent tasks and large-scale nodes.The application comprises the following steps: dividing a large-scale constellation into constellation management domains, determining the shortest domain-level path from a starting constellation management domain to a terminal constellation management domain for each task, determining the initial entry domain endpoint and the initial exit domain endpoint of each constellation management domain based on a global routing reference direction, adjusting the exit domain endpoint of a congested node to obtain a final exit domain endpoint and a corresponding final entry domain endpoint, and dividing the tasks in the constellation management domain into rounds according to the estimated time sequence of the tasks to arrive at the constellation management domain, and selecting a low-load next-hop satellite node for each round of task.The application reduces the complexity of constellation routing planning under large-scale concurrent tasks and large-scale nodes, and realizes the optimization of network load balancing performance.

This application discloses a lightweight, anti-blocking intelligent routing method, apparatus, device, medium, and product for satellite constellations based on Stackelberg game theory, relating to the field of satellite communication networks. The method includes: constructing a satellite constellation network simulation environment and establishing a Stackelberg game model, defining the blocker as the leader and satellites as followers; designing a hierarchical routing framework, including a routing selection module based on an Actor-Critic architecture and an anti-blocking module based on Q-Learning; updating routing network parameters using a near-end policy optimization algorithm and utilizing Q-Learning to quickly reselect paths after detecting blockage; and achieving algorithm lightweighting through artificial dimensionality reduction of the state space and knowledge distillation techniques. This application effectively improves the robustness and reliability of satellite constellation networks in intelligent blocking environments, realizes a lightweight design of the intelligent algorithm, proves the existence of Stackelberg equilibrium, and significantly improves routing success rate and reduces average latency.

The application discloses a wireless sensor network clustering routing based on an improved beluga whale optimization algorithm and relates to the field of Internet of Things.The technical scheme of the application improves the beluga whale optimization algorithm through student t distribution and cosine boundary strategy, which is called tCBWO.BWO algorithm is in the local development stage most of the time during the algorithm execution process; a DPR heuristic algorithm is proposed and applied to WSN; a new hierarchical routing scheme of WSN is proposed.The stability of the BWO algorithm is improved, and the Prim algorithm is improved in combination with a transmission energy consumption model of WSN, and is named as tCBWO-DPR algorithm.The algorithm can provide high-quality data transmission service in each round and can effectively improve the balance of the network.

This invention proposes a continuous learning method and evaluation method to mitigate the double forgetting of visual language models. The method includes: adding two parallel paths to the backbone network of a pre-trained visual language model—one a task-agnostic expert path and the other a task-related expert group path—to obtain a model based on heterogeneous expert hybridization; training the model using a two-stage training strategy, including: a first stage, freezing the task-related expert group path and training only the task-agnostic expert path, using contrastive learning to solidify the pre-trained general knowledge of the visual language model; a second stage, freezing the task-agnostic expert path and training only the task-related expert group path, using a cross-entropy loss function to learn task-specific knowledge; and an inference stage, dynamically fusing the outputs of the two parallel paths through a hierarchical routing mechanism to achieve collaborative learning of pre-trained knowledge and task-specific knowledge.

This invention relates to a hierarchical adaptive federated learning method, apparatus, and medium. The method includes: a server initializing global model parameters and broadcasting them to a client; a client initializing local model parameters; the client extracting scaling factors for each batch normalization (BN) layer in the local model and calculating hierarchical channel importance scores to generate a personalized channel set and a shared channel set; performing hierarchical routing fusion of local and global parameters based on the personalized and shared channel sets to update the local model parameters; training the local model and applying adaptive constraint losses to the parameters corresponding to the personalized and shared channel sets during training; after training, the client pruning and differentially privacy-processing the updated amount of the shared parameters based on the BN layer scaling factors and uploading it to the server; the server updating the global shared parameters and distributing them to the client; repeating the aforementioned steps for a preset number of rounds to obtain the local model.

The application discloses a kind of multi-task data analysis systems based on hybrid expert model, comprising: data input module is standardized to multi-task original data;Task feature extraction module extracts the corresponding feature expression data in multi-task data set;Hierarchical gated routing module determines the expert path of task using hierarchical routing algorithm;Hierarchical expert network module executes feature fusion according to routing result;Multi-task prediction module executes task prediction to the fused features;Multi-task loss optimization module optimizes model parameters based on gradient normalization and uncertainty weighting algorithm;Model parameter update module realizes parameter synchronous update;Multi-task result output module carries out final correction of prediction result.The application effectively improves the stability and accuracy of multi-task data analysis.