17results about How to "Improve computing power" patented technology

This invention discloses a liquid-cooled heat dissipation chip structure and its packaging method, relating to the field of chip technology. The liquid-cooled heat dissipation chip structure includes: a chip with accommodating channels formed therein, and coolant filled into the accommodating channels; a substrate, on which the chip is packaged; and a cooling fan mounted on the side of the chip away from the substrate. This invention solves the problem of chip overheating leading to decreased chip computing performance and slowed data processing speed.

The application provides an activation module for data calculation, an AI processor and a data calculation method. When it is determined that a result calculated by a preceding matrix acceleration module meets an output condition, the output calculation data is stored in a cache unit. A read-write control unit reads the output calculation data from the cache unit and distributes the output calculation data to each calculation unit. Each calculation unit performs a softmax calculation and outputs a target calculation result. In this way, the output calculation data output by the preceding matrix acceleration module can be stored in the cache unit of the activation module. After processing by the read-write control unit and data distribution, the softmax calculation is performed in parallel by at least one calculation unit. The calculation and access of all data are performed in the activation module, reducing the overhead of accessing memory and improving overall calculation performance. The parallel calculation processing scheme can also improve calculation efficiency.

The application provides a synthetic aperture radar active composite jamming method and system, the method comprising: acquiring intermediate frequency radar signals to be jammed, radar related parameters and control parameters; calling DSP and multiple FPGA calculation boards to calculate modulation coefficients corresponding to each false scattering point according to pre-downloaded deception templates and radar related parameters; generating deception jamming signals corresponding to the intermediate frequency radar signals according to the intermediate frequency radar signals and the modulation coefficients corresponding to each false scattering point; generating suppression jamming signals corresponding to the intermediate frequency radar signals according to the control parameters; and determining composite jamming signals corresponding to the intermediate frequency radar signals according to the deception jamming signals and the suppression jamming signals corresponding to the intermediate frequency radar signals. The application can obtain composite jamming signals, has better jamming effect, is not easy to expose the position of the jammer, and has lower cost of the jammer. The signal processing architecture of DSP and FPGA is adopted, the calculation performance of the system is improved, and the real-time performance of system simulation is ensured.

This invention discloses a computation graph optimization method, system, and storage medium based on a multi-level cost model, comprising: (1) defining a multi-level cost model to evaluate the execution cost of computation graph execution schemes; (2) setting a set of subgraph replacement rules; (3) converting a given neural network model file into unified computation graph information; (4) analyzing the candidate computation graph structure and applying the subgraph replacement rules to obtain a new computation graph; (5) selecting a cost model to evaluate the execution performance of the computation graph; (6) selecting candidate computation graphs based on the cost evaluation results and scaling factors; and (7) repeating steps (4) to (6) until the expected computation graph optimization effect or convergence condition is achieved to obtain the optimal computation graph. This invention comprehensively considers the trade-off between search time and optimization effect, can meet different needs in various scenarios, guide the optimization direction of computation graphs, and thus improve the computational efficiency and performance of neural network models.

This invention discloses an autonomous evolutionary AGI system and method based on dual-model decoupling and human feedback, belonging to the field of artificial intelligence technology. The system includes modules for online inference (Model A), offline training (Model B), human feedback rewards, persistent storage, training scheduling, model hot replacement, distributed communication, and a secure sandbox. This invention decouples inference and training to ensure service continuity; anchors value direction with human ratings to avoid drift; employs dual-condition triggering training to improve resource efficiency; achieves seamless updates through hot replacement; avoids memory leaks through database persistence; enables multi-machine expansion through distributed architecture; and ensures execution security through a sandbox. This invention enables 24 / 7 online, controllable, and secure autonomous evolution of AGI, suitable for code generation, intelligent agents, automated R&D, and general artificial intelligence scenarios.

This application relates to a cache management method, apparatus, computer device, medium, chip, and product. The method includes: acquiring a target task and a management command queue; the management command queue containing various management commands for the target task; when the last-level cache module does not contain dependency data required by a target subtask within the target task, in response to a first management command, reading the dependency data required by the target subtask from memory and storing the dependency data in the last-level cache module; in response to a second management command, initiating the execution of the target subtask; the target subtask includes an accelerator reading dependency data from the last-level cache module. This method can improve the processing efficiency of the target task.

The application discloses a kind of computing resource scheduling method and system based on computing power network cluster, and the application relates to the technical field of computing power network cluster, according to the amount of calculation of multiple computing tasks, the node load state of multiple computing power nodes and resource scheduling system Determine the computing task matched by multiple computing power nodes, Ensure the matching accuracy of multiple computing power nodes to computing task.Therefore, according to the computing progress of multiple computing power nodes to computing task and the computing mode of multiple computing power nodes Determine the computing progress distribution diagram of computing power network cluster;According to the computing progress distribution diagram of computing power network cluster, the computing resource matched by each computing power node and the task state of multiple computing tasks Determine the optimization system of computing resource, and dynamically optimize the computing power of computing power node in computing process, based on the optimization system of computing resource Realize the further optimization of computing resource, improve the computing power of computing power node in computing process.

The application relates to a localized multi-network-port high-computing-power data acquisition device, which comprises a core module and an interface expansion module, wherein the core module and the interface expansion module are connected; the core module is based on a Ruikai micro rk3588 platform; the interface expansion module comprises five independent gigabit network ports, two USB-A interfaces, one M.2 high-speed storage interface, one Type-c high-speed interface, one HDMI output interface, one Mini-Pcie expansion interface and a plurality of expansion IO interfaces; the structure provided in the application has the advantages that the computing power is improved: the heterogeneous architecture of RK3588 NPU+extension acceleration card, the AI task processing efficiency is improved by more than 5 times compared with a traditional industrial computer; the multi-network-port redundancy: 4-network-port design supports protocol isolation and link backup, and the network availability reaches 99.99%; localization adaptation: the core chip, the peripheral chip and resistance, capacitance and inductance are all domestic, and the component localization rate is 100%; flexible expansion: 5G, LoRa and other communication modules can be expanded through the mini-PCIe and USB interfaces, and the device is suitable for diversified scenes.

The application discloses a data operation device and method, a processing core and an electronic device. The device comprises a data reading module, which is used for receiving an instruction and reading a first matrix and a second matrix based on the instruction. The instruction comprises grouping parameters of a calculation unit array. The grouping parameters are parameters used for dividing the calculation unit array into sub-calculation arrays. The grouping parameters are related to rows of the first matrix or columns of the second matrix. The sub-calculation arrays read data of the first matrix and the second matrix and perform operations of the first matrix and the second matrix. According to the grouping parameters in the instruction, the data operation device divides the calculation unit array into multiple sub-calculation arrays, which can realize flexible combination of calculation units of the calculation unit array, and can effectively utilize the calculation units to improve the computing power of the data operation device.

This application provides a resource scheduling method, apparatus, electronic device, medium, and program product, relating to the application of large-scale models in fintech scenarios, and applicable to the fields of big data technology and artificial intelligence technology. The method includes: acquiring multi-source heterogeneous data, wherein the multi-source heterogeneous data includes at least customer-side, environment-side, and system-side data; constructing a multimodal feature representation based on the multi-source heterogeneous data; inputting the multimodal feature representation into a customer behavior prediction model to perform multi-task prediction, obtaining prediction state information; obtaining resource allocation strategy information based on the prediction state information and resource scheduling objectives; and performing resource scheduling based on the resource allocation strategy information.

This application provides a method, system, and device for hierarchical forwarding of vehicle-road-cloud data based on edge nodes, relating to the field of intelligent transportation technology. The method includes: real-time collection of vehicle-road-cloud traffic data through an integrated vehicle-road-cloud system, and deployment of N edge computing nodes on the roadside; feature extraction and data analysis of the vehicle-road-cloud traffic data based on the N edge computing nodes; hierarchical processing of the vehicle-road-cloud traffic data according to the analysis results; construction of a vehicle-road-cloud data forwarding strategy library; differential analysis of the hierarchical vehicle-road-cloud data based on the library; determination of the target data hierarchical forwarding strategy; and hierarchical forwarding control of the vehicle-road-cloud traffic data. This application solves the technical problems of low data processing efficiency caused by excessively high data transmission latency, uneven network bandwidth allocation, and inefficient computation in existing technologies. By deploying edge computing nodes on the roadside, it improves the processing efficiency and real-time performance of traffic data.

This invention relates to a smart glasses with better weight distribution and privacy protection, comprising a head-mounted support element, a rotatable lens support on one side of the support element, and a waveguide lens for displaying digital content mounted on the lens support. This invention utilizes the head to support the lens support and the waveguide lens. This structural design minimizes the discomfort and even pain caused by prolonged wear on the bridge of the nose. Furthermore, the physical rotation of the lens support provides an intuitive indication of the camera's on / off status, addressing concerns about privacy leaks.

This invention proposes an efficient K-short path calculation method and system for dynamic road networks, relating to the field of path planning technology. It involves obtaining multiple OD pairs from a directed graph of the road network, constructing a list, and initializing it. For each OD pair, a shortest path is calculated, and a set of deviation paths is constructed based on the shortest path. The shortest path and the set of deviation paths are stored as an initial K-short path set. When the edge weights of the road network structure are adjusted, potentially affected OD pairs are screened based on the changed edge attributes and the initial K-short path set. The initial K-short paths from the potentially affected OD pairs are obtained, and a substantial impact test is performed: if an initial K-short path does not contain any changed edges, and its path cost is less than the total cost of potential new paths constructed from the changed edges, then the path is retained; otherwise, the path is deleted. New paths are supplemented based on the set of deviation paths until K shortest paths are satisfied, thereby significantly reducing computational complexity and improving the efficiency and accuracy of path updates in dynamic scenarios.

This invention discloses a machine learning feature extraction method based on parallel quantum convolutional neural networks (QNNs). It converts a batch of "data blocks" into a qubit system, enabling the simultaneous processing of multiple data blocks in a single round of QNN computation. Furthermore, all data blocks within the same batch share the same parameterized quantum gates, reducing the number of free circuit parameters while maintaining the locality of spatial features. This invention significantly improves the computational efficiency, accuracy, and robustness of quantum convolutional neural networks.

This invention discloses a traveling wave protection method, device, system, and medium based on heterogeneous computing, belonging to the field of power system relay protection. The method is based on a heterogeneous computing architecture that coordinates CPU and FPGA, executing high-speed data sampling, phase-mode transformation, start-up discrimination, and integral calculation in parallel via the FPGA; the CPU is responsible for parameter configuration, receiving preprocessing results from the FPGA, and executing the final protection logic judgment. Specifically, the start-up discrimination uses an integer arithmetic formula optimized for FPGA to quickly capture faults; the integral calculation obtains the integral value of the first-mode current within a specific time window; the CPU calculates its correlation with the reference waveform using a discrimination function to distinguish between faults inside and outside the fault zone. This invention significantly improves the computational efficiency and operating speed of traveling wave protection, exhibiting high reliability and strong tolerance to transition resistance.

This invention proposes a real-time simulation system and method for multi-energy network digital twins based on fully embedded systems. It develops digital twins of various energy networks based on real-time simulation using a fully embedded method, and proposes fully embedded models and convergence radius models for various energy networks. The convergence radius model is used to obtain convergence information of the fully embedded model. The convergence radius model is determined by simplifying the power series coefficients of the unbalance function of the time-varying dynamic process of the multi-energy network state variables, thereby enhancing the computational performance of the fully embedded model and achieving real-time standards.