35results about How to "Solve training problems" patented technology

The invention discloses a block chain consensus method based on deep learning training tasks, including that all block chain participating nodes jointly maintain a ledger record, any node can issue a transaction, and distribute it to each node through a P2P consensus network, There will be deep learning task nodes in the P2P consensus network, and the deep learning tasks will be distributed to all participating nodes. The accounting nodes will combine the deep learning tasks to reach a consensus through the inter-node consensus mechanism and complete the transaction confirmation. The present invention combines artificial intelligence and deep learning technology, adds the workload proof of the block chain public chain mining process into the deep learning task, effectively solves the traditional energy waste problem of obtaining bookkeeping rights through hash challenges, and to a certain extent It solves the problem of deep learning model training; at the same time, the workload proof of the mining process retains part of the challenge using the hash algorithm SHA256, which ensures the randomness of the mining process and avoids the monopoly of bookkeeping rights by deep learning giants.

The invention discloses a neuromorphic system based on variable-resistance devices and adaptive-excited neurons and a realization method. The system includes a crossed array of the variable-resistance devices, the anterior neurons, the posterior neurons, a global dynamic threshold value control circuit, a control logic module, a voltage regulation module, a sample input, a tag input and a result output. According to the system, the variable-resistance devices are used as electronic synapses, and a new structure and operation mode of the adaptive-excited neurons are provided, thus the system is optimized on area and operation aspects, and training problems faced by the same type of systems are solved.

The invention discloses a simulation method for generating a three-dimensional scene based on a Unity dynamic technology, belonging to the field of simulation of a communication scene of a power system. The method comprises the following steps of step 1, constructing three-dimensional models, more specifically, three-dimensional models of a communication machine room and communication equipment, wherein the communication equipment consists of device sub-racks, single plates and optical fibers; step 2, newly constructing a C# procedure in Unity to read a user-defined configuration file and automatically generating a three-dimensional scene simulation module; step 3, newly constructing a monitoring thread in the three-dimensional scene simulation module, subscribing an information channel of a real-time database, and obtaining external real-time data required for generating the three-dimensional scene simulation module from an external equipment simulation procedure; and step 4, operating the equipment or repairing an equipment failure in the three-dimensional scene, and transmitting the operation information to the external equipment simulation procedure to carry out simulating calculation and judgment. According to the method, the operating status of the equipment in the simulation module of the communication equipment can be displayed in real time, and operations of failure checking and failure repairing can be flexibly carried out.

The invention discloses a method and system for re-identifying pedestrians who change clothes based on an autoencoder network, wherein the method includes: (1) generating a mask for the pedestrian's clothing part in the pedestrian picture; (2) using a clothing feature encoder E A Obtain the clothes feature representation vector; (3) remove the clothes part in the pedestrian picture according to the mask, and use the clothes changing picture generator G to generate the clothes changing picture; (4) build a clothes-independent feature learning network, including the feature extractor F and the picture generation (5) pair the real pedestrian pictures and the generated clothes-changing pictures in pairs, and train the feature extractor F; (6) after the training converges, input the real and marked pedestrian pictures, and use the cross-entropy loss function to The feature extractor F is fine-tuned; (7) The application of pedestrian re-identification is carried out by using the final feature extractor F. Utilizing the present invention, robust discriminative features can be learned in the scene where the appearance of pedestrians varies widely.

The invention relates to a substation training device based on a PLC (programmable logic controller). The substation training device comprises a base, an operating display screen, an alternating current air switch, an alternating current contactor, an intermediate relay, the PLC, a grounding fault setting device, a wiring terminal block, PLC signal input devices and PLC signal output devices, wherein the base is provided with a power plug, and the operating display screen comprises a box shell, a simulation circuit board, a warning device, an isolation switch operating rocking rod, a displayer I, a displayer II and a plurality of voltmeters. The substation training device has the advantages that through controlling the operating display screen by the PLC, one-time analogue simulation training of a substation is realized, the training problem of electricity utilization operating personnel is solved, the educational and training quality of trainees is fast promoted, the acceptability of electricians for network access operation is greatly improved, the operation is visual and vivid, the trainees can easily grasp the learned key points, a notable effect is produced, and the problem that the actual operation is very complex and is difficult to grasp is solved.

The invention provides a method for improving the activity of nitrification function microbiology in activated sludge through directly feeding Fe ions, which comprises a cultivation stage of the nitrification function microbiology and a long-time running activity maintenance stage. In a saprobiont denitrification system, ferric salt chemical agent with Fe <2+> or Fe<3+> is directly fed into an activated sludge mixture of an aerobic pond of a biology denitrification system, and basic compounds are also fed simultaneously; and in the long-time running activity maintenance stage of nitrificationfunction microbiology, the feeding amount of the ferric salt chemical agent with the Fe <2+> or the Fe<3+> is gradually reduced, and the fed ferric salt chemical agent with the Fe <2+> or the Fe<3+> is only used for replenishing loss amount of the Fe<3+> or the Fe <2+>. According to the method, the biochemical reaction metabolic activity of the nitrification function microbiology can be effectively improved, the breeding of the nitrification function microbiology and the secretion of endoenzyme of microbial cell are promoted, the applicability of saprobiont denitrification technology is improved, the bottleneck problems existing in the saprobiont denitrification system are solved, and the low temperature resistant capacity of the system is obviously improved.

The invention discloses an image super-resolution method based on a deep threshold convolutional neural network, which is mainly used to solve the problem that the super-resolution effect of an image is reduced when the network is deepened in the prior art. The method comprises the following steps: (1) acquiring paired low-resolution and high-resolution image data as training data; (2), defining a threshold convolution layer to replace the existing convolution layer, and building an end-to-end deep threshold convolutional neural network; (3) inputting the training data to the deep threshold convolutional neural network to train the network through an Adam optimization method; and (4) using the trained deep threshold convolutional neural network to carry out image super-resolution. The gradient attenuation problem of the deep neural network is reduced. The application of the deep network in image super resolution is realized. The effect of image super-resolution is enhanced. The speed of image super-resolution is increased. The image super-resolution method can be used in fields such as satellite remote sensing, medicine, traffic monitoring and video compression.