Abstract: Disclosed are a quantization method and quantization apparatus for a weight of a neural network, and a storage medium. The neural network is implemented on the basis of a crossbar-enabled analog computing-in-memory (CACIM) system, and the quantization method includes: acquiring a distribution characteristic of a weight; and determining, according to the distribution characteristic of the weight, an initial quantization parameter for quantizing the weight to reduce a quantization error in quantizing the weight. The quantization method provided by the embodiments of the present disclosure does not pre-define the quantization method used, but determines the quantization parameter used for quantizing the weight according to the distribution characteristic of the weight to reduce the quantization error, so that the effect of the neural network model is better under the same mapping overhead, and the mapping overhead is smaller under the same effect of the neural network model.

Abstract: A hierarchical audio-visual feature fusing method for audio-visual question answering and a product relate to the field of audio-visual question answering. By fusing audio embedding in an input video clip with a baseline model as well as video embedding and question embedding respectively at an early stage, a middle stage and a late stage in a hierarchical feature fusing process, a first answer probability distribution, a second answer probability distribution and a third answer probability distribution are obtained, and the answer probability distributions are added based on preset weights, and then averaged for hierarchical integration to generate a final answer.

Abstract: An infrared detector includes a thermoelectric element, an infrared light absorber located on the thermoelectric element, and an electrical signal detecting element. The infrared light absorber includes a first drawn carbon nanotube film, a second drawn carbon nanotube film, and a third drawn carbon nanotube film stacked on each other. The first drawn carbon nanotube film includes a plurality of first carbon nanotubes substantially extending along a first direction. The second drawn carbon nanotube film includes a plurality of second carbon nanotubes substantially extending along a second direction. The third drawn carbon nanotube film includes a plurality of third carbon nanotubes substantially extending along a third direction. The first direction and the second direction form an angle of about 42 degrees to about 48 degrees, and the first direction and the third direction form an angle of about 84 degrees to about 96 degrees.

Abstract: The present disclosure provides a self-driven water collecting surface having a superhydrophilic-superhydrophobic structure, and a method for preparing the same, belonged to the technical field of water harvesting and superhydrophobic surfaces. The water collecting surface is a superhydrophobic surface with the distributed superhydrophilic region. The superhydrophilic region is a venation channel network structure consisting of hierarchical superhydrophilic channels. In the method, a pulsed laser is firstly adopted to form periodically distributed peak-pit microstructures and nanostructures, which is then modified with a low-surface-energy substance. Then, the low-surface-energy substance layer is removed by a pulsed laser again according to a venation channel network pattern. The laser scanned region is superhydrophilic, while the other regions are superhydrophobic. So that, the self-driven water collecting surface with the superhydrophobic-superhydrophilic structure is obtained.

Abstract: Embodiments of the present disclosure provide a Bayesian image denoising method based on a distribution constraint of noiseless images from a distribution of noisy images. The method includes: determining a distribution constraint of noiseless images from a distribution of noisy images; and performing Bayesian denoising on noisy images based on the distribution constraint of noiseless images.

Abstract: A method of multi-directional scene text recognition based on multi-element attention mechanism include: performing normalization processing for a text row/column image output from an external text detection module by a feature extractor, extracting a feature for the normalized image by using a deep convolutional neural network to acquire an initial feature map, adding a 2-dimensional directional positional encoding P to the initial feature map in order to output a multi-channel feature map, converting the multi-channel feature map output from a feature extractor by an encoder into a hidden representation, and converting the hidden representation output from the encoder into a recognized text by a decoder and using the recognized text as the output result.

Abstract: An image collection chip, an object imaging recognition device and an object imaging recognition method are provided. The image collection chip comprises an optical modulation layer and an image sensing layer. The optical modulation layer is located on the image sensing layer. The optical modulation layer is provided with at least two modulation units and the image sensing layer is provided with sensing units corresponding to the at least two modulation units in a form of up and down. Each of the at least two modulation units is provided with at least one modulation subunit. Two or more modulation units among the at least two modulation units have different graphic structures, and the two or more modulation units having different graphic structures have different modulation roles on spectrum.

Abstract: A modularized parametric visual program induction algorithm, a device, a medium and a product are provided. An optimized modularized parametric model is obtained by constructing a modularized parametric model and training with the cooperation of a hierarchical Monto-Carlo Tree-Search algorithm, data to be processed is input into the optimized modularized parametric model and processed in conjunction with the hierarchical Monto-Carlo Tree-Search algorithm, and a result program expression is output. The modularized parametric visual program induction algorithm is applicable to complex visual scenes; and model training data of the modularized parametric model is augmented in a training stage in conjunction with the hierarchical Monto-Carlo Tree-Search algorithm.

Abstract: The present disclosure relates to an analog-to-digital conversion circuit comprising: N sampling and conversion modules connected in parallel, configured to simultaneously sample and sequentially convert first analog signals of N channels to output second analog signals, wherein each of the sampling and conversion modules includes a plurality of sampling capacitors connected in parallel, wherein N is an integer greater than 1; a comparator connected to the N sampling and conversion modules, configured to comparing the second analog signals respectively to obtain comparison signals; and a control module connected to the N sampling and conversion modules and the comparator, configured to control the N sampling and conversion modules to output converted digital signals based on the comparison signals.

Abstract: An device for measuring electron beam comprises a Faraday cup comprising an opening; a porous carbon material layer located on a surface of the Faraday cup and suspended at the opening; and an electricity meter electrically connected to the porous carbon material layer. A length of a suspended portion of the porous carbon material layer is greater than or equal to a maximum diameter of an electron beam to be measured. A diameter or a width of the porous carbon material layer is smaller than a minimum diameter of a cross section of the electron beam to be measured. The porous carbon material layer comprises a plurality of carbon material particles, and a plurality of micro gaps exist between the plurality of carbon material particles. A method for measuring an electron beam using the device for measuring electron beam is also provided.

Abstract: An architecture search method and an architecture search apparatus for a large-scale graph, and a device and a storage medium are provided. The method includes: obtaining a subgraph of a large-scale graph by performing local sampling on the large-scale graph; sampling a plurality of neural network architectures in a pre-constructed super network according to a pre-customized importance sampling strategy; obtaining a plurality of trained neural network architectures by training, according to a peer learning method, the plurality of neural network architectures with the subgraph; obtaining a trained super network by iteratively executing the subgraph sampling, the architecture sampling, and the architecture training; and selecting an optimal architecture in the trained super network to process the large-scale graph to obtain a graph processing result.

Abstract: A method includes: pre-processing initial data of an acquired to-be-analyzed confrontation scenario, to obtain triplet data marked in a graph data structure, inputting the triplet data into a pre-trained analyzing model, to determine a result of analysis on semantic information of the to-be-analyzed confrontation scenario; wherein the analyzing model includes a plurality of operator networks that are provided with execution priorities, wherein the plurality of operator networks are configured for analyzing relations of different types; and the processing result of any one of the operator networks includes: in the relation type corresponding to the operator network, respective true relations of all of the node pairs, and confidences and descriptive values of the true relations; and according to the graph data structure that is updated by using the processing results of all of the operator networks, characterizing the result of analysis on the semantic information of the to-be-analyzed confrontation scenario.

Abstract: An automatically tray feeding device includes: at least one conveying mechanism configured to convey a tray at different heights; at least one lifting mechanism configured to move the tray on a conveying mechanism at a first height to a conveying mechanism at a second height different from the first height; and a sorting table frame configured to carry the at least one conveying mechanism and the at least one lifting mechanism.

Abstract: The disclosure discloses an electro-hydraulic servo actuator capable of implementing long-stroke and high-frequency loading and a control method. The actuator includes: a long hydraulic cylinder with a piston, a short hydraulic cylinder with a piston, a fixed-end rod, low frequency and high frequency electro-hydraulic servo valves, a left rod chamber, a right rod chamber, a force applying end head, and a force applying end rod. With this structure, the displacement of the piston of the short hydraulic cylinder is a combination of small-displacement and high-frequency movement and large-displacement and low-frequency movement, so as to realize the generation of large-displacement and high-frequency movement. Therefore, a target large-displacement and high-frequency excitation is applied to the specimen under load test. The actuator can output high-frequency and large-displacement target excitation of 0.

Abstract: A preparation method of a perfluoropolyether-based magnetic liquid includes dispersing magnetic nanoparticles coated with graphene oxide into a solution of N,N-dimethylformamide to obtain solution A; dispersing a surfactant into dichloromethane, and adding triethylamine as a cosolvent to obtain solution B; mixing the solution A and the solution B uniformly, heating them under reflux and stirring them for a modification reaction; and after the modification reaction is completed, washing and drying a reaction product to obtain the modified magnetic nanoparticles coated with the graphene oxide; dispersing the modified magnetic nanoparticles coated with the graphene oxide into a base carrier liquid to prepare the perfluoropolyether-based magnetic liquid. The surfactant is a perfluoroalkylamine. The base carrier liquid is a perfluoropolyether oil. The modification reaction is performed at a temperature of 50 to 120° C. for a time period of 20 to 50 hours.

Abstract: The present disclosure provides a software protection method and apparatus, an electronic device and a storage medium. By identifying the sensitive data in the software source code and determining the point of use of the sensitive data; setting the sensitive data in the read-only section of the compilation product of the software source code so that the sensitive data is stored in the read-only area when the software is running; inserting the check code into the above point of use of the sensitive data, where the check code is used to check whether the sensitive data read is in the read-only area while the software is running, and determine whether to continue running the software according to the check result, embodiments of the present disclosure realize the integrity verification of the sensitive data, and thereby prevent the sensitive data from being corrupted with low performance overheads.

Abstract: The present disclosure provides a smart lock, including: a lock body and an unlocking structure; a Bluetooth communication module configured to receive a lock instruction or an unlock instruction based on Bluetooth communication; and one or more binding tags, each binding tag being associated with a Bluetooth address of the smart lock, wherein the unlocking structure is configured to lock or unlock the smart lock based on the lock instruction or the unlock instruction, or lock or unlock the smart lock based on an operation performed for the one or more binding tags. The present disclosure further provides a system and method for monitoring an object equipped with the smart lock.

Abstract: A fixing clamp for a microfluidic chip includes a set of sub-clamps arranged to be spaced apart on left and right. Each of the sub-clamps includes an upper plate body with a first passage, the first passage having a first outer interface and a first chip docking port, and the first chip docking port being located on a lower surface of the upper plate body and at one end of the upper plate body facing the other sub-clamp; a lower plate body with a second passage, the second passage having a second outer interface and a second chip docking port, and the second chip docking port being located on an upper surface of the lower plate body and at one end of the lower plate body facing the other sub-clamp; and a spacing adjusting mechanism connecting the upper plate body and the lower plate body together.

Abstract: The present disclosure provides an anisotropic lamellar inorganic fiber aerogel material and a preparation method thereof. The method includes: mixing a polymer solution, an inorganic precursor and a chloride to obtain a spinning precursor solution; blow spinning the spinning precursor solution to obtain a composite fiber aerogel; calcinating the composite fiber aerogel to obtain the anisotropic lamellar inorganic fiber aerogel material. Therefore, the method has advantages of simplicity, easy operation, low cost, high efficiency and easy industrialized production. The inorganic fiber aerogel materials prepared by the above method are composed of multi-layer stacked fibers and have an anisotropic lamellar structure, which can be cut into any desired shape, and stacked to any desired thickness.

Abstract: An organic electroluminescent device and a display apparatus. The organic electroluminescent device includes a light-emitting layer, including a triplet-triplet annihilation type host and a fluorescent dye, where the fluorescent dye has a structure represented by the following Formula (1) or Formula (2). When the fluorescent dye in the light-emitting layer of the organic electroluminescent device combines with the triplet-triplet annihilation type host, the voltage of the device can be reduced, and the luminescence efficiency of the device can be improved.