Abstract: The present disclosure relates to a tensor-based optimization method for GPU memory management of deep learning, at least comprising steps of: executing at least one computing operation, which gets tensors as input and generates tensors as output; when one said computing operation is executed, tracking access information of the tensors, and setting up a memory management optimization decision based on the access information, during a first iteration of training, performing memory swapping operations passively between a CPU memory and a GPU memory so as to obtain the access information about the tensors regarding a complete iteration; according to the obtained access information about the tensors regarding the complete iteration, setting up a memory management optimization decision; and in a successive iteration, dynamically adjusting the set optimization decision of memory management according to operational feedbacks.

Abstract: A method for preparing a non-corrosive fire extinguishing agent, including combining a methyl phosphate organosilane material and a perfluorohexanone to prepare the fire extinguishing agent.

Abstract: The disclosure belongs to the technical field related to on-line measurement in manufacture of integrated circuit, which discloses a snapshot type overlay error measuring device and a measuring method thereof. The measuring method includes: the detection light is subjected to polarization and retardation in sequence to obtain measurement spectrum; Fourier analysis is performed on the measurement spectrum to obtain the frequency-domain signal of the measurement spectrum, and sub-channel frequency-domain analysis is performed on the frequency-domain signal to obtain the linear combination of the non-diagonal Mueller matrix elements of the overlay error sample to be tested; the linear combination of the non-diagonal Mueller matrix elements are processed to obtain the overlay error of the overlay sample under test. This disclosure does not need to measure all 16 Mueller matrix elements, the measurement is carried out on only a few non-diagonal Mueller matrix elements which are sensitive to overlay error.

Abstract: The disclosure belongs to the technical field of solid fuel utilization and discloses a gasification reactor adaptable for feedstock with wide particle size distribution, including a reactor body. The reactor body is composed of a first reaction chamber, a second reaction chamber, and a third reaction chamber, which are connected with each other. The side wall of the first reaction chamber is provided with a first vent for introducing a gasification agent to fluidize the fine feedstock particles in the first reaction chamber and the gasification reaction occurs. The bottom of the second reaction chamber is provided with a second vent for introducing an oxidant to react with the coarse feedstock particles in the second reaction chamber. The bottom of the third reaction chamber is provided with a third vent for introducing a gasification agent to fluidize and gasify the incompletely reacted particles in the third reaction chamber.

Abstract: A phase change memory device based on a nano current channel is provided. A nano current channel layer structure is adopted and configured to limit the current channel. As such, when flowing through the layer, the current enters the phase change layer from nano crystal grains with high electrical conductivity, and the current is thereby confined in the nano current channels. By using the nano-scale conductive channels, the contact area between the phase change layer and the electrode layer is significantly decreased, the current density at local contact channel is significantly increased, and heat generation efficiency of the current in the phase change layer is improved. Moreover, an electrically insulating and heat-insulating material with low electrical conductivity and low thermal conductivity prevents heat in the phase change layer from being dissipated to the electrode layer, and Joule heat utilization efficiency of the phase change layer is thereby improved.

Abstract: Disclosed in the present invention are a chalcogenide phase change material based all-optical switch and a manufacturing method therefor, relating to the field of optical communications. The all-optical switch comprises: stacked in sequence, a cover layer film, a chalcogenide phase change material film, an isolation layer film, a silicon photonic crystal, and a substrate. The silicon photonic crystal comprises a nano-porous structure such that the silicon photonic crystal has a Fano resonance effect. When the all-optical switch is used, the state of the chalcogenide phase change material film is controlled by means of laser, and the resonance state of the silicon photonic crystal is modulated to implement modulation of signal light transmissivity; the modulation range is within a communication band from 1500 nm to 1600 nm, thereby implementing an optical switch. The all-optical switch of the present invention has the characteristics of high contrast ratio, high rate and low loss.

Abstract: The present disclosure relates to an FPGA-based dynamic graph processing method, comprising: where graph mirrors of a dynamic graph that have successive timestamps define an increment therebetween, a pre-processing module dividing the graph mirror having the latter timestamp into at least one path unit in a manner that incremental computing for any vertex only depends on a preorder vertex of that vertex; an FPGA processing module storing at least two said path units into an on-chip memory directly linked to threads in a manner that every thread unit is able to process the path unit independently; the thread unit determining an increment value between the successive timestamps of the preorder vertex while updating a state value of the preorder vertex, and transferring the increment value to a succeeding vertex adjacent to the preorder vertex in a transfer direction determined by the path unit, so as to update the state value of the succeeding vertex.

Abstract: A method for synthesizing an aliphatic polyester block copolymer regulated by carbon monoxide. The method uses an organic cobalt metal complex as a convertible catalyst and includes: first regulating an anionic ring-opening copolymerization reaction between an aliphatic acid anhydride and an epoxy compound; then using carbon monoxide as a conversion agent for conversion to obtain a catalyst having a new catalytic active site; and regulating a vinyl monomer to perform active free radical polymerization to obtain an aliphatic polyester block copolymer having a controllable structure. The described synthesis method uses a convertible catalyst and a conversion agent, combines two controllable polymerization reactions which have different but compatible mechanisms, and obtains a block copolymer by means of a “one-pot” reaction.

Abstract: The disclosure belongs to the field of quench protection of a superconducting magnet system and specifically relates to a quench protection circuit for a superconducting magnet system based on a distributed heater network including M superconducting coils connected in series and a heater network formed by N heater modules, where M>N and N?3. Different heater modules are connected in parallel with different superconducting coil subsets, and all superconducting coil subsets have spatial symmetry. Each heater module has m parallel branches, and each parallel branch has n heaters connected in parallel, where m?1, n?1, and when N=1, m>1. Each heater in the heater network is thermally coupled to one superconducting coil among the M superconducting coils, and each superconducting coil is thermally coupled to at least one heater in each heater module.

Abstract: The present disclosure belongs to the field of materials with metal structures, and specifically relates to a preparation method for a nano-oxide dispersion strengthened steel. The method includes mixing a ferrochromium alloy, a ferrotungsten alloy, a ferroalloy containing a rare earth element, an oxygen source and a reduced iron powder to obtain a mixture; wrapping the mixture in a steel strip, and conducting drawing reducing to obtain a flux-cored wire; and conducting arc additive manufacturing on the flux-cored wire on a substrate, and then conducting heat treatment to obtain the nano-oxide particle dispersion strengthened steel.

Abstract: The present invention belongs to the technical field of simulation of power semiconductor modules, and discloses a multi-physics co-simulation method of a power semiconductor module. The multi-physics co-simulation method of the power semiconductor module comprises: adopting professional circuit simulation software PSpice supporting a spice model to be imported into a device, and by designing a specific collaborative analysis method and performing secondary development of a software data exchange interface, i.e. constructing a coupling interface of co-simulation, performing electricity-heat-force co-simulation of two types of software PSpice and COMSOL by adopting an indirect coupling manner. The simulation time is greatly shortened, and the simulation efficiency is improved.

Abstract: The present invention relates to a Docker-container-oriented method for isolation of file system resources, which allocates host file system resources according to access requests from containers and checks lock resources corresponding to the access requests. The method creates a plurality of new containers; allocating the host file system resources according to file resource request parameters required by the new containers; and controlling execution of the file system operation according to an amount of the file system resources that have been allocated to the new containers.

Abstract: A method and a system for correcting lithography process hotspots based on stress damping adjustment are provided. The method includes: acquiring a mark hotspot of a mask pattern; forming N annuli centered on the mark hotspot from inner to outer on a mask; moving vertexes of the mask pattern located in each annulus by a specific distance in a direction deviating from the mark hotspot and connecting the moved vertexes according to an original connection relationship to acquire an updated layout; verifying electrical characteristics of the updated layout, determining whether a deviation of the electrical characteristics of the updated layout is within a tolerable range, and performing geometric correction to compensate for a deviation of electrical parameters if no is determined and then ending correction, or ending the correction if yes is determined.

Abstract: The invention belongs to the field of amorphous alloys, and more specifically, relates to a method for calibrating the internal temperature field of amorphous alloy prepared by spark plasma sintering. First, the part required for temperature field calibration inside the bulk amorphous alloy sample obtained by spark plasma sintering is cut into a series of small amorphous alloy samples, and the isothermal crystallization treatment is performed to obtain the crystallization time of different parts of the sample. An annealing-isothermal crystallization experiment is performed on the adopted amorphous alloy powder at different annealing temperatures, and the functional relationship between the annealing temperature and the crystallization time is obtained.

Abstract: A multi-modal field-programmable metamorphic sensor is provided. The metamorphic sensor has a multi-layer structure, including an upper electrode layer, a spacer layer, and a lower electrode layer. At least one of the upper electrode layer and the lower electrode layer is a sensor A. The spacer layer is disposed between the electrode layers. A sensor B is formed through forming a structure of electrode layer-spacer layer-electrode layer structure. When detecting an object, the measurement of different physical quantities of different objects is implemented through switching between the sensor A and the sensor B. The sensor serves as a single cell. Multiple single cells are combined to form a multi-cell sensor. A signal acquisition system for acquiring signal of the sensor is also provided.

Abstract: A deep learning heterogeneous computing method based on layer-wide memory allocation, at least comprises steps of: traversing a neural network model so as to acquire a training operational sequence and a number of layers L thereof; calculating a memory room R1 required by data involved in operation at the ith layer of the neural network model under a double-buffer configuration, where 1?i?L; altering a layer structure of the ith layer and updating the training operational sequence; distributing all the data across a memory room of the CPU and the memory room of the GPU according to a data placement method; performing iterative computation at each said layer successively based on the training operational sequence so as to complete neural network training.

Abstract: The disclosure provides a dynamic interaction-oriented subject's limb time-varying stiffness identification method and device. The method includes: the combination of subject's limb displacement and measured force data or the combination of angle and measured torque data is collected; based on the time-varying dynamic system constructed based on a second-order impedance model, the linear parameter varying method is utilized to substitute the time-varying impedance parameters and reconstruct the restoring force/torque expression; iterative identification is performed on variable weights, dynamic interaction force/torque, and restoring force/torque by using time-varying dynamic parameters based on the dynamic interaction force/torque expression expanded from basis function; the time-varying stiffness is solved by using variable weights and dynamic interaction force/torque according to expression with substituted the time-varying impedance parameters.

Abstract: The present invention belongs to the technical field related to additive manufacturing, and provides a multi-field composite-based additive manufacturing device and method. The device comprises a powder delivery adjustment module, a sound field control module, a microwave field/thermal field control module and a microprocessor.

Abstract: The present invention relates to a container-based network function virtualization (NFV) platform, comprising at least one master node and at least one slave node, the master node is configured to, based on interference awareness, assign container-based network functions (NFs) in a master-slave-model-based, distributed computing system that has at least two slave nodes to each said slave node in a manner that relations among characteristics of the to-be-assigned NFs, info of load flows of the to-be-assigned NFs, communication overheads between the individual slave nodes, processing performance inside individual slave nodes, and load statuses inside individual said slave nodes are measured.

Abstract: A method for configuring a reconfigurable physical unclonable function (PUF) based on a device with spin-orbit torque (SOT) effect is provided. The disclosure uses SOT or magnetic field to change the magnetic moment. After the current or magnetic field is removed, the magnetic moment returns to the easy axis direction. Under the effect of thermal fluctuation, the magnetic moment is randomly oriented in the easy axis direction. The non-volatile devices are formed into an array, the magnetic moments of all non-volatile devices are randomly distributed after a write operation. The read state can be used as a random code to implement the reconfigurable PUF. The PUF has a simple structure and guarantees security. The random code in the disclosure may be two-state or multi-state, which is related to the number of magnetic domains of the ferromagnetic layer. A large number of challenge response pairs form a strong PUF.