Abstract: An LED display screen, comprising: an LED array, consisting of multiple LED light-emitting units and used for emitting a light; an optical diffusion film, provided at a light exit side of the LED array; a matrix shading frame, comprising multiple hollow shading gratings, the hollow shading gratings corresponding one-to-one to the LED light-emitting units; and a substrate, used for supporting the LED array and the matrix shading frame, where the light emitted by the LED light-emitting units, after running through the hollow shading gratings, is diffused to a viewer side via the optical diffusion film, and the LED light-emitting units emit the light towards the hollow shading gratings.

Abstract: A high-speed remote landslide simulation test device with a variable angle includes a support adjustment component, slide plates, and loose leaves. The device first transmits the operation paths of adjusting the slide plate at different angles to the controller, and the personnel operates the control panel to control the support jack, the motor, and the electromagnet to start and stop through the controller, the output shaft of the motor drives the stainless steel threaded rod to rotate, the controller first energizes one electromagnet and disconnects the other three electromagnets, which can only make the stainless steel threaded rod rotate in a fixed position inside the inner threaded pipe, and the three inner threaded pipes follow the rotation direction of the stainless steel threaded rod, thus, the position of the support jack can be moved separately.

Abstract: The present disclosure provides a blockchain-based green power certification method, apparatus, and system, to pre-process obtained basic data to obtain proof-of-existence data, certify the proof-of-existence data from a production side and a consumption side by using a certification chain and a token chain respectively, and obtain a green power attribute certificate and a green power consumption label if the certification is passed; generate a unique digital identifier for the green power attribute certificate and the green power consumption label separately based on a non-fungible token (NFT), which realizes non-interchangeability, indivisibility, and other characteristics of the certificate and the identifier; and supervise and verify a certification process by using a chain of custody, and generate a corresponding unique blockchain identifier for the certification chain, the token chain, and the chain of custody separately through multi-chain interaction, so as to realize data interaction between blockcha

Abstract: Embodiments of this disclosure include a data processing method. In the method, a training sample set that includes a plurality of graph computing task training samples is obtained. At least one performance indicator feature of each of the graph computing task training samples is extracted. The at least one performance indicator feature includes one or more of a graph data feature, a graph processing platform feature, a graph algorithm feature, and a machine hardware feature. A target performance prediction model is generated based on a mapping relationship between actual execution times of the graph computing task training samples and the performance indicator features. According to at least one performance indicator feature of an inputted graph computing task test sample, a predicted execution time of the graph computing task test sample is output based on the target performance prediction model.

Abstract: A fatigue bending and folding test device for an ultra-thin metal strip, including: a first driver module, a first folding module, a first diameter adjustment module, an expansion module, and a frame. The first driver module, the first folding module, the first diameter adjustment module, and the expansion module are all disposed on the frame. The first driver module includes a servo motor, a coupler, and a drive gearbox; the drive gearbox includes a driving gear shaft, a driven gear shaft, a plurality of output end covers, and a plurality of bearings. The first folding module includes an upper folding member, a lower folding member, and a first mounting base. The first clamping member includes a sliding block, a lower clamping plate, an upper clamping plate, a rotating shaft, an intermediate pressure plate, and a plurality of fixing bolts.

Abstract: A data processing method is provided. In the method, a historical walk vertex adjacent to a target walk vertex is determined. An edge transition probability between the target walk vertex and each of a set of next possible vertexes in a first out-neighbor set is determined according to first out-edge information. A to-be-reached vertex of the set of next possible vertexes in the first out-neighbor set is determined according to the edge transition probabilities. Second out-edge information corresponding to the target walk vertex is generated based on the first out-neighbor set. Walking from the target walk vertex to the to-be-reached vertex is performed. The second out-edge information is transmitted to the to-be-reached vertex. Further, a random walk sequence corresponding to the target walk vertex is generated based on a walk quantity corresponding to the target walk vertex reaching a preset threshold for walk steps.

Abstract: A data processing method is provided. In the method, a historical walk vertex adjacent to a target walk vertex is determined. An edge transition probability between the target walk vertex and each of a set of next possible vertexes in a first out-neighbor set is determined according to first out-edge information. A to-be-reached vertex of the set of next possible vertexes in the first out-neighbor set is determined according to the edge transition probabilities. Second out-edge information corresponding to the target walk vertex is generated based on the first out-neighbor set. Walking from the target walk vertex to the to-be-reached vertex is performed. The second out-edge information is transmitted to the to-be-reached vertex. Further, a random walk sequence corresponding to the target walk vertex is generated based on a walk quantity corresponding to the target walk vertex reaching a preset threshold for walk steps.

Abstract: Embodiments of this disclosure include a data processing method. In the method, a training sample set that includes a plurality of graph computing task training samples is obtained. At least one performance indicator feature of each of the graph computing task training samples is extracted. The at least one performance indicator feature includes one or more of a graph data feature, a graph processing platform feature, a graph algorithm feature, and a machine hardware feature. A target performance prediction model is generated based on a mapping relationship between actual execution times of the graph computing task training samples and the performance indicator features. According to at least one performance indicator feature of an inputted graph computing task test sample, a predicted execution time of the graph computing task test sample is output based on the target performance prediction model.

Abstract: The present invention relates to artificial neural networks, for example, deep neural networks. In particular, the present invention relates to a compression method for deep neural networks with proper use of mask and the device thereof. More specifically, the present invention relates to how to compress dense neural networks into sparse neural networks while maintaining or even improving the accuracy of the neural networks after compression.

Abstract: A multi-iteration method for compressing a deep neural network into a sparse neural network without degrading the accuracy is disclosed herein. In an example, the method includes determining a respective initial compression ratio for each of a plurality of matrices characterizing the weights between the neurons of the neural network, compressing each of the plurality of matrices based on the respective initial compression ratio, so as to obtain a compressed neural network, and fine-tuning the compressed neural network.

Abstract: A packet classification method includes building a decision tree with a preset building method for a packet, extracting packet header bits indicated by the bitmask of current node and concatenating the bits to generate a child index, traversing the child nodes recursively according to the child index until the leaf node is reached, obtaining in the reached leaf node a list of rule pointers referring to rules, and matching each of the rules so as to classify the packet. The method improves upon the current trade-off, and achieves faster classification speed while retaining reasonable memory consumption. A device for packet classification can implement the method.

Abstract: The present invention relates to artificial neural networks, for example, deep neural networks. In particular, the present invention relates to a multi-iteration compression method for deep neural networks and the device thereof.

Abstract: The present invention relates to artificial neural networks, for example, deep neural networks. In particular, the present invention relates to a compression method for deep neural networks with proper use of mask and the device thereof. More specifically, the present invention relates to how to compress dense neural networks into sparse neural networks while maintaining or even improving the accuracy of the neural networks after compression.

Abstract: Sudden cardiac arrest is treated by reducing blood temperature from about 37° C. to 33° C., following resuscitation, by injecting hypothermia inducing drugs such as a cannabinoid type into the patient's body, preferably in combination with physical surface body cooling.

Abstract: Sudden cardiac arrest is treated by reducing blood temperature from about 37° C. to 33° C., following resuscitation, by injecting hypothermia inducing drugs such as a cannabinoid type into the patient's body, preferably in combination with physical surface body cooling.