23results about How to "Reduce learning difficulty" patented technology

The invention discloses a photovoltaic power prediction method based on improved empirical mode decomposition and an optimized long short-term memory network, and the method comprises the steps: firstly carrying out the preprocessing of abnormal value elimination, missing value filling, normalization and the like of photovoltaic power and related meteorological data, and improving the data quality; then, an improved empirical mode decomposition (EE-ANEMD) algorithm is adopted to decompose the preprocessed power sequence into a multi-scale intrinsic mode function component and a residual term, and high-frequency noise, intermediate-frequency fluctuation and a low-frequency trend are effectively separated; global optimization is carried out on the hidden layer unit number, the initial learning rate and the maximum number of training times of the LSTM network through an improved sparrow search algorithm (ISSA), finally, the optimized LSTM is utilized to carry out training prediction on each component, and results are fused and subjected to reverse normalization to obtain a final value. Experiments show that the test set RMSE of the method is reduced compared with that of a single LSTM, the mid-term prediction precision is remarkably improved, and reliable technical support is provided for power system dispatching, new energy consumption planning and photovoltaic power station operation and maintenance.

This invention discloses a face forgery detection method based on identity decoupling and adaptive cosine embedding loss. Addressing the problems of existing deepfake detection methods being susceptible to interference from identity information and lacking flexibility due to fixed boundary constraints, this invention constructs an identity-independent deepfake detection network. It decouples the input face image into forensic features containing forgery clues and content features containing identity information, eliminating identity interference by ignoring content features. Simultaneously, it designs an adaptive cosine embedding loss based on visual differences, using absolute error and the visual differences between real and fake image pairs as adaptive boundary values ​​to dynamically adjust feature spacing constraints. This invention significantly improves the model's generalization ability when facing unknown forgery types through the synergistic optimization of decoupled representation learning and adaptive metric learning.

The application is suitable for the field of artificial intelligence technology, and provides a word vector representation method and device based on language processing and a terminal device, the method comprising: constructing an input word table; setting word frequency information of each word in the input word table; classifying the words into N clusters according to the word frequency information, words with the same word frequency information being classified into the same set to obtain a clustering set with N subsets; defining a first word vector dimension of the words in the first cluster through an experience factor, and obtaining a second word vector dimension of the words in each subset according to the subset where the words are located and the first word vector dimension; defining an nth linear mapping matrix for the nth cluster; and processing a target word vector through parameters shared by the input word table as an input of a natural language processing model. Through the application, the problem that the calculation amount is large, the efficiency is low and the word distribution features cannot be considered when converting words into word vectors in natural language processing can be solved.

The specification provides a language model training method, device and program product. The training process of the language model is divided into multiple learning stages. The language model training method comprises: generating multiple constraint conditions for original question samples; based on the original question samples and the corresponding multiple constraint conditions, constructing multiple constraint question samples containing different numbers of constraint conditions; in each learning stage, generating a training set based on all the constructed constraint question samples; each learning stage corresponds to a target number, and the target number increases accordingly as the learning stage progresses; in the training set, the proportion of constraint question samples with the target number of constraint conditions is higher than that of other constraint question samples; training the current language model of each learning stage based on the training set of the learning stage; wherein, from the second learning stage, the current language model of any learning stage is the language model trained in the previous learning stage.

This disclosure provides a method for deploying a neural network model, an inference method, and an electronic device. The deployment method includes: determining the chip platform corresponding to the neural network model; compiling a platform-level dynamic library and a device-level dynamic library corresponding to the chip platform, wherein the device-level dynamic library is configured to provide a first set of interfaces to the platform-level dynamic library, and the first set of interfaces is configured to call hardware-related interfaces of the chip platform; the platform-level dynamic library is configured to provide a second set of interfaces to the application layer, and the second set of interfaces is configured to call the first set of interfaces of the device-level dynamic library; and deploying the neural network model, the platform-level dynamic library, and the device-level dynamic library to the chip platform. This disclosure improves model deployment efficiency.

This utility model discloses an automatic testing system for flight controllers, relating to the field of unmanned aerial vehicle (UAV) technology. It includes a control unit, a sensor unit, a storage unit, a detection unit, a power supply unit, a display unit, an input unit, and an interface unit. The control power supply includes a master MCU and a slave MCU. The control unit is connected to the sensor unit, storage unit, detection unit, power supply unit, display unit, input unit, and interface unit. The sensor unit includes a compass, a barometer, and an IMU, and is connected to the detection unit. The storage unit contains a TF card. This utility model eliminates the need for additional instruments, reducing the cost of purchasing equipment. Furthermore, only the automatic testing system needs to be learned, reducing the learning difficulty and manual training costs for workers. Simultaneously, the flight controller is connected to the system via a single cable, reducing connection difficulty and shortening testing time from approximately 2 hours to about 10 minutes.

The invention provides an energy consumption prediction method, device and system for an air conditioning system. The accuracy and reliability of energy consumption prediction can be improved. The method comprises the steps of obtaining real-time operation state data of an air conditioning system of a building; a data driving model is obtained, the data driving model is obtained through training on the basis of historical operation state data and a physical mechanism model of the air conditioning system, and a neural network structure adopted by the data driving model is a long-short-term memory network; the physical mechanism model is constructed based on static physical parameters of the building, historical environmental data and a preset thermodynamic law; and the real-time operation state data is processed based on the data driving model, an energy consumption predicted value is obtained, and the energy consumption predicted value is used for generating a control instruction for adjusting the air conditioning system.

The invention relates to the field of surgical instruments, in particular to a handheld control device and a multi-degree-of-freedom surgical instrument. The deflection mechanism is rotationally connected to the holding body and provided with a first driving column and a second driving column, and the deflection mechanism can be shifted to drive the first driving column and the second driving column to deflect relative to the holding body; the transmission mechanism is connected to the first driving column and the second driving column so as to drive the first driving column and the second driving column to rotate. According to the device, the independent deflection mechanism and transmission mechanism are arranged, so that physical decoupling of two main actions of opening, closing and deflection of the tail end clamp body is realized; deflection and rotation are mutually independent, control paths are clearly separated through overall deflection or respective rotation of the driving column, the motion coupling phenomenon caused by linkage of driving wires in the prior art can be avoided, control is more accurate, the hand feeling is direct, extra compensation by a doctor is not needed, operation complexity is reduced, and operation efficiency is improved. And the continuity and the accuracy of the operation are ensured.

The present application relates to the technical field of image processing, and more particularly to a priori knowledge guided wavelet-space double attention image rain removal method and system, comprising: obtaining an original rainfall image as an initial input image; using a priori knowledge guided wavelet-space double attention rain removal deep learning network model to perform rain removal processing on the initial input image to obtain a rain-free image; the a priori knowledge guided wavelet-space double attention rain removal deep learning network model comprises an input mapping module, a priori knowledge guiding module, a wavelet-space double attention based hierarchical encoding-decoding network and an output mapping module. The rain removal deep learning network model used in the present application combines a priori knowledge guiding mechanism, wavelet-space double attention and multi-scale feature interaction strategy to enable the network model to better understand the structure and features of rainy day images, enhance feature extraction, improve image detail and texture quality, and reduce network learning difficulty.

The invention discloses a power Internet of Things terminal access parameter adaptive optimization method and system based on hierarchical reinforcement learning, and the method comprises the steps: carrying out the optimization modeling of a terminal access parameter into a hierarchical reinforcement learning decision process, and carrying out the access strategy selection and parameter fine-grained adjustment through a high-layer intelligent agent and a low-layer intelligent agent, and in combination with Monte Carlo sampling, simulation evaluation is carried out on a dynamic network environment, so that online adaptive optimization of access parameters of the terminal equipment is realized. In consideration of diversity and complexity faced by equipment access in an electric power internet of things system, the access parameters of the equipment are optimized in real time by dynamically analyzing the running state and service requirements of the terminal equipment and combining five modules including a sampler, a parameter judgment device, an object model state machine, a benchmark tester and a reward model. By adopting the method provided by the invention, challenges such as terminal large-scale access and heterogeneous network environment dynamic change in the power internet of things can be effectively handled, and the terminal access reliability, the network throughput and the communication efficiency are remarkably improved.

The application relates to the technical field of medical teaching, in particular to a proximal humerus fracture Neer classification teaching system and method, wherein the system comprises: a data mapping module, which is used for automatically identifying a finger state and analyzing the finger state into a fracture displacement parameter of corresponding proximal humerus anatomical structure according to a mapping relationship; and a rendering simulation module, which is used for driving a three-dimensional proximal humerus fracture model according to the fracture displacement parameter, superimposing the three-dimensional proximal humerus fracture model on the fingers or presenting the three-dimensional proximal humerus fracture model in a virtual space for rendering simulation in a visual environment. The application simulates fracture displacement according to a mapping relationship and a gesture rule by establishing a mapping relationship between the fingers and the anatomical structure, converts abstract medical classification into physical actions that can be personally perceived and operated by learners, and thus greatly reduces learning difficulty and improves understanding depth and memory persistence.

This utility model belongs to the technical field of casting equipment, and relates to an arc-shaped plate transverse stacked box casting structure, including an upper sand box and a lower sand box that cooperate with each other. The lower sand box has a horizontal sprue on its parting surface. The upper sand box contains a first sand core and several second sand cores stacked together. Each second sand core has a casting cavity for forming the casting, and a lettering core is placed within the casting cavity to form a mark on the casting. The second sand core has an ingate that connects to the horizontal sprue. The upper sand box also has a sprue that connects to the horizontal sprue. A pouring cup is placed on the top surface of the upper sand box, and the outlet of the pouring cup connects to the upper inlet of the sprue. This utility model is easy for workers to operate, has a low learning curve, saves effort, improves production efficiency, is safe to use, reduces safety hazards, and effectively solves the problem of low yield caused by one-piece casting per box process, thus improving production efficiency.

This invention discloses a macrophage digestion solution and a method specifically for macrophage digestion, comprising the following components: 3-6 parts ethylenediaminetetraacetic acid (EDTA), 4-8 parts ethylene glycol bis(2-aminoethyl ether)tetraacetic acid (EDTA), 8-12 parts fetal bovine serum (FBS), and 90-100 parts phosphate buffer. The method specifically for macrophage digestion includes the following steps: preparing a stock solution of digestion solution, filtering it, and storing it at low temperature for 6 months to obtain the macrophage digestion solution; adding the macrophage digestion solution to a macrophage culture flask; shaking the macrophage culture flask containing the digestion solution to ensure even coverage; adding DMEM culture medium to the settled macrophage culture flask to neutralize the digestion solution, and transferring it to a centrifuge tube; centrifuging the solution in the centrifuge tube at room temperature for 3 minutes; and adding DMEM culture medium to the centrifuge tube after centrifugation to complete the macrophage digestion. This invention achieves high macrophage digestibility and low differentiation rate.

This invention relates to the field of intelligent decision-making technology, and discloses a search result ranking method, apparatus, device, and medium, comprising: receiving a search request sent by a user terminal and determining the corresponding search object; acquiring the first initial profile features of the current user, the behavioral features of the current user clicking on the search object, and the second initial profile features of all search objects; converting the continuous variables in the first and second initial profile features into dispersed variables to obtain the first target profile features and the second target profile features, respectively; calculating the click probability of each search object based on the first target profile features, behavioral features, second target profile features, and current search object features using a trained search prediction model; sorting the search objects according to their click probabilities from high to low, and selecting a specified number of the top-ranked search objects as search results. This method can accurately calculate the click rate of search objects and has good search result prediction capabilities.

This utility model relates to a simple geometry learning aid device, comprising two wooden boards; the two wooden boards are hinged together, and a digital angle ruler is installed on the front side of one of the wooden boards. At least two horizontal tubes are rotatably connected to the upper surfaces of both wooden boards. A scale is provided on the lower surface of each wooden board, and a groove is formed inside the scale along its length. Two fixed components are slidably connected inside the groove. A common pull rope is connected to the outer sides of the two fixed components, and a common zipper is also provided on the outer sides of the two fixed components. This simple geometry learning aid device can simultaneously demonstrate the teaching of plane analytic geometry and solid geometry, transforming abstract geometric knowledge into intuitive graphics and operable processes. It helps students better understand the definitions and characteristics of conic sections such as circles, ellipses, and hyperbolas, as well as the relationships between lines, lines and planes, and planes in solid geometry, reducing learning difficulty and improving learning efficiency.

The utility model relates to medical instrument technical field, concretely is a unilateral double -channel endoscope operation lock target cavity expansion pipe, including two -level expansion sleeve, three -level expansion sleeve, four -level expansion sleeve, five -level expansion sleeve, six -level expansion sleeve and two primary guide rod, two -level expansion sleeve, three -level expansion sleeve and four -level expansion sleeve both ends are equipped with first empty groove in symmetrical position, five -level expansion sleeve and six -level expansion sleeve one side have the second empty groove with two -level expansion sleeve, three -level expansion sleeve and four -level expansion sleeve equal width, and this expansion pipe passes through the design of grooving, and the tip of primary guide rod of observation channel can be embedded in the slot mouth, and the meeting target point of locking observation channel and operation channel is avoided to avoid target point displacement, solves the problem of repeated positioning in operation, and reduces the risk of iatrogenic accidental injury.

The application discloses a machine learning assisted phase shift mask proximity correction method and belongs to the technical field of semiconductor lithography. The method comprises the following steps: obtaining an initial phase shift mask pattern and a target imaging pattern; constructing and training a physically guided forward imaging prediction model, wherein the model adopts a hybrid neural-physical double-branch architecture, the physical branch uses a learnable parameterized effective thin mask to approximately compensate for the thick mask effect, the main branch learns the residual error, and the fusion output predicts the spatial image; taking the forward model as a differentiable simulator, performing reverse correction under the manufacturability constraint by using the accompanying sensitivity and controlled random disturbance, and obtaining a final phase shift mask pattern; and outputting the mask pattern data after design rule checking and local fine-tuning. The application takes into account the physical consistency and data-driven flexibility of imaging prediction, improves the reverse optimization quality, considers the mask manufacturability constraint, and is suitable for proximity correction of phase shift masks in EUV lithography.

The application provides an inner ring outer spherical surface grinding micron-level precision control device and a use method, and belongs to the technical field of machining lathes; the device comprises a machine tool cover and a machine tool base, and the bottom of the machine tool cover is fixedly connected with the machine tool base. The polishing module is arranged, the module can not only polish the workpiece surface and adsorb and clean the debris at the same time, but also can reduce the probability of damage of the workpiece caused by the debris not being cleaned in time during the polishing process to a certain extent, thereby assisting in guaranteeing the yield of the product to a certain extent, the workpiece does not need to exit the polishing head to be cleaned by water flushing, which saves the workpiece processing time to a certain extent and assists in improving the workpiece processing efficiency, the debris accumulation degree of the module is proportional to the wear of the polishing abrasive wheel piece, and therefore the module can also trigger a replacement of the polishing abrasive wheel piece after the debris accumulates to a certain degree, which can assist in guaranteeing the polishing effect of the polishing head to a certain extent.

The present application relates to a new type of electronic music pipe, which is innovatively designed on the basis of traditional electronic blowing pipe, characterized by a blow nozzle-free structure, which completely changes the playing method and sanitary conditions, significantly reduces the requirement for breath control, and greatly widens the range of applicable people; the button layout is designed rigorously, the note buttons and auxiliary buttons are distributed reasonably, the layout of the left and right hand areas is convenient for players with different habits to operate, the note buttons are integrated with a force sensor, which is installed at the bottom of the button or below the circuit board, senses the playing force, and further realizes multi-dimensional control of the volume and timbre of the music, and gives the player rich expression; it has professional karaoke function, creates diverse musical atmosphere with the help of advanced audio synthesizer and multi-channel output instrument chip; the internal function corresponding table is open, and the functions of the buttons and voice commands can be personalized set; the electronic music pipe reduces the learning threshold, enriches the music performance, promotes music innovation and technological progress, and widens the application scenarios.

The utility model relates to a kind of casting pouring structure with small hole, including the upper box and lower box of being connected, mud core assembly and pouring structure are equipped in upper box and lower box, and the casting cavity between upper box casting cavity, lower box casting cavity, mud core assembly is casting cavity, and pouring structure communicates casting cavity;Pouring structure includes air outlet and sequentially communicated straight sprue, cross gate, inner gate, straight sprue upper end communicates sprue cup, and inner gate communicates casting cavity bottom, and casting cavity top communicates air outlet;Mud core assembly includes the casting mud core and the mould release mud core of symmetrical arrangement, small hole core bracing component is installed on casting mud core, small hole core bracing component includes the core head and small hole core bracing rod of being connected, and small hole core bracing rod other end is connected with casting mud core;Small hole core bracing rod middle is small hole core bracing part.The utility model effectively solves the problem that the shape of casting is affected by the fracture of sand core in weak wall thickness due to low strength.

The application discloses a kind of automatic wharf horizontal transport equipment dirty detection method and device, construct first level convolutional neural network for detecting IGV from image and second level convolutional neural network for detecting dirty area from IGV, first level convolutional neural network and second level convolutional neural network are constructed into double-branch cascade detection model, the original image obtained from the automatic wharf operation scene is input as the double-branch cascade detection model, and the first level convolutional neural network identifies IGV from complex scene, and outputs the image labeled IGV boundary box, according to the boundary box, IGV subgraph containing only IGV image is cut out from image, so as to remove background interference and transmit to second level convolutional neural network, and second level convolutional neural network focuses on detecting dirty area from IGV subgraph, so as to improve the detection efficiency and detection precision of IGV dirty area.

The present application relates to the field of medical devices, and provides an ablation catheter and an ablation system comprising the same, the ablation catheter comprising a penetrating part, an ablation main body, a catheter main body and a control handle connected in sequence; the penetrating part comprises a screwing-in member capable of screwing into and being fixed to a target tissue under the action of an external force; the ablation main body comprises an ablation electrode, a connecting member and a temperature sensor; the catheter main body comprises a non-ablation electrode and a perfusion channel; the non-ablation electrode is arranged in a spaced manner with the ablation electrode, and the non-ablation electrode is not in contact with the target tissue during ablation. By using the ablation catheter and the ablation system, the thickened interventricular septum is screwed and ablated, the screwing speed is easier to control, the depth can be stopped or adjusted in time before penetrating the myocardial wall, myocardial perforation can be prevented, and the damage to the myocardial tissue is reduced through temperature detection and perfusion function.

This invention relates to a progressive fusion method for medical images based on learning how to optimize paradigms, belonging to the field of medical image processing technology. It aims to solve the problems of difficult optimization, poor modal adaptability, and low inference efficiency in traditional end-to-end fusion methods. The method first converts multimodal medical images from RGB to YCbCr color space and extracts the luminance component. An initial fusion image is generated by projecting the pixel-level maximum value, and an initial triplet state is constructed and stored in a memory playback pool. A fusion flow estimation network samples the state from the memory pool to predict pixel change flow. After iteratively training the network, the trained network is used to progressively optimize the initial fusion image. Finally, the chrominance components are concatenated and converted back to RGB space to obtain the fusion result. This invention models fusion as a dynamic trajectory optimization problem, combining the memory playback mechanism and flow matching theory, reducing the learning difficulty of the network, mitigating differences between modal domains, achieving controllable fusion direction and smooth, continuous trajectory. Accurate fusion can be completed with very few iterations, balancing fusion accuracy, model generalization ability, and inference efficiency, meeting the high efficiency and reliability requirements of clinical diagnosis.