ZHEJIANG LAB

An event prediction method based on a satellite orbit threat domain knowledge graph is provided. By constructing a knowledge graph of the satellite orbit threat domain, and utilizing steps such as meta-path extraction, hyperedge construction, feature encoding, and feature aggregation, target features that can comprehensively and accurately describe each entity and the complex relationships between each entity in the satellite orbit threat domain are obtained, and a prediction model is trained by using the target features of each entity.

The present application discloses a fiber optic sensor and a measurement method, apparatus, and storage medium. By performing, based on the microwave signal, intensity modulation on the optical signal, then inputting the modulated optical signal into the fiber optic containing the weak reflection grating array, a reflected signal is obtained. Thereafter by performing dispersion compensation on the amplified partial reflected signal, and based on the dispersion-compensated reflected signal and the non-dispersion-compensated reflected signal, the change amounts respectively corresponding to positions in the fiber optic containing the weak reflection grating array are determined.

An asymmetric cross-modal large-model knowledge transfer method for remote sensing includes: acquiring a training sample pairs including a sample RGB image and a sample MS image corresponding to a same scene classification; inputting the sample MS image into a teacher model; determining a first image feature extracted by the teacher model from the sample MS image; determining a first scene classification, obtained by the teacher model according to the first image feature, as a pseudo label; inputting the sample RGB image into a student model; determining a second image feature extracted by the student model from the sample RGB image; determining a second scene classification obtained by the student model according to the second image feature; and training the student model according to a difference between the second image feature and the first image feature and a difference between the second scene classification and the pseudo label.

The present application discloses a fiber optic sensor and a measurement method, apparatus, and storage medium. By performing, based on the microwave signal, intensity modulation on the optical signal, then inputting the modulated optical signal into the fiber optic containing the weak reflection grating array, a reflected signal is obtained. Thereafter by performing dispersion compensation on the amplified partial reflected signal, and based on the dispersion-compensated reflected signal and the non-dispersion-compensated reflected signal, the change amounts respectively corresponding to positions in the fiber optic containing the weak reflection grating array are determined.

An information propagation prediction method based on a sequential hypergraph neural network with co-attention fusion. The method includes three modules: a user feature learning module learning a user embedding by utilizing a graph convolutional neural network; a cascade feature learning module constructing a sequential hypergraph neural network based on equivariant diffusion to capture a complex cascade irregular connection so as to learn a cascade embedding of internal and external features of an encapsulated cascade structure; and a feature fusion and prediction module learning an interdependence between cascade features and user features by means of a co-attention mechanism so as to capture a complex relationship and interaction between the cascade features and the user features, and then calculating a possibility of potential user infection. The present disclosure achieves more effective information propagation prediction, comprehensively considers the external influence of other cascades, and combines a novel feature fusion technology.

A method for detecting network attack based on kernel operating characteristics of a software switch, comprising: collecting kernel operating characteristic data of the software switch during network management and packet forwarding of the software switch, wherein the kernel operating characteristic data of the software switch can reflect characteristic data of a fine-grained operating state of a network; preprocessing the kernel operating characteristic data of the software switch; and inputting the preprocessed kernel operating characteristic data of the software switch into a pre-trained attack detection model to detect a potential attack behavior in a network. The method is applicable to all scenarios where the software switch is applicable, especially a software-defined network and a cloud data center, an industrial Internet, 5G, edge computing based on the software-defined network. More accurate and timely detection and early warning of potential anomalies and attack behaviors in the network can be implemented.

Provided are a satellite network construction method and apparatus, a medium, and a device. For a satellite in a satellite cluster, candidate satellites of the satellite are determined based on orbit data of the satellite, and communication quality between the satellite and the candidate satellites within first time in the future is predicted. Based on the predicted communication quality, a resource quantity of each of the candidate satellites, and the orbit data, networking satellites matching the satellite are determined from the candidate satellites to construct a satellite network of the satellite. A to-be-executed distributed task is executed in the satellite cluster in units of the satellite network. Another satellites with good quality of communication with the satellite are determined as the networking satellites of the satellite to construct the satellite network of the satellite. Then, the distributed task is executed in units of the satellite network.

The present invention discloses a two-dimensional photonic convolutional acceleration system and device for convolutional neural network, comprising: a multi-wavelength light source, a signal source to be convolved, a modulator, a dispersion module, a 1×M power divider, an optical fiber delay array, a microring weighting array chip, a convolutional kernel matrix control unit, a trans-impedance amplifier array, and an acquisition and processing unit. The present invention realizes two-dimensional convolutional acceleration based on wavelength-time interleaving technology, a single modulator can realize the optical domain loading of the signal, and the convolutional operation speed is only limited to the speed of the modulator. The present invention can realize two-dimensional convolutional kernel convolutional acceleration of two-dimensional data in a single signal cycle based on two-level delay and microring weighting array chip, and solve the problem of data redundancy in traditional methods, the scheme is simple and efficient. The present invention realizes the control of convolutional kernel matrix coefficient based on the microring weighting array chip, can realize the fast update of convolutional kernel matrix coefficient, and is suitable for real-time data processing applications, the balanced photodetector can realize arbitrary convolutional kernel coefficient weighting.

A method and device for interpreting a graph neural network based on FPGA acceleration propose to use FPGA hardware to accelerate interpretation process of the graph neural network oriented to node classification in parallel, and improve node traversal and shortest path search of BFS, thereby optimizing requirements of algorithm calculation and storage, and accelerating generation of interpretation results. During calculating HN values, the present disclosure optimizes multiplication operation using the matrix characteristics, transforms the dense matrix multiplication into sparse-dense matrix multiplication, and optimizes the resource occupation using multi-PE parallel processing, greatly improving performance of graph neural network interpretation acceleration. Moreover, an overall architecture based on FIFO storage calculation task distribution is designed to reduce calculation difference between nodes, solving the difficulty of high time complexity of the graph neural network interpretation method based on node classification in actual data applications and improving the time efficiency of interpretation.

The present disclosure discloses a method and an apparatus for data processing, a storage medium and an electronic device, where for each second unit in the embodiments of the present disclosure, a plurality of communication links are provided between the first unit and the second unit. The first unit, in response to a data operation request, sends the data operation request to the second unit through the plurality of communication links between the first unit and the second unit. The second unit processes the target data to be processed according to the data operation instruction in the data operation request to obtain a processing result, and sends the processing result to the first unit via a plurality of communication links. The first unit executes a response strategy for the data operation request based on the processing result.