16results about How to "Mitigate performance degradation" patented technology

The invention discloses a heterogeneous federal fine tuning language model construction method and system based on differential privacy, and the method comprises the steps: a parameter server firstly carries out the clustering of clients with similar capabilities, carries out the sensing clustering according to resources, obtains a larger-rank adaptive matrix for each class according to the class with the large computing power of the computing resources, and carries out the calculation of the resources; therefore, the flexible rank of the low-rank adaptive matrix is realized, and the fine tuning performance is enhanced. In each class, privacy federal fine tuning is adopted, including two parts of noise distribution with sensitivity perception and regular term optimization, so that better training efficiency and reduced noise influence are obtained. Between each class, heterogeneous low rank adaptive matrix aggregation is performed through a knowledge distillation technique, including knowledge aggregation and selection of high quality knowledge to enhance the local model. According to the method, strict privacy guarantee can be provided, the influence of differential privacy on fine tuning of the low-rank adaptive matrix can be relieved in a heterogeneous environment, and higher efficiency and accuracy are realized.

The invention discloses a preparation method of a vesuvianite coarse aggregate anti-crack Portland cement-based material, which comprises the following steps: (1) soaking vesuvianite coarse aggregate in a saturated ammonium dihydrogen phosphate solution, and drying to obtain pretreated aggregate; (2) uniformly mixing carbon powder with a solution containing a saturated magnesium ion source, adding ammonia water, and pyrolyzing the obtained mixture to obtain pretreated carbon powder; and (3) uniformly mixing the pretreated carbon powder with molten stearic acid, heating, preserving heat, cooling to room temperature, and grinding to obtain the anti-cracking agent. And (4) adding the anti-cracking agent into the pretreated aggregate of which the surface is wetted by an anhydrous solvent under a stirring condition, uniformly mixing, heating, preserving heat, and cooling to room temperature to obtain the modified coarse aggregate. And (5) uniformly mixing Portland cement, the modified coarse aggregate, the fine aggregate, the filler, the silica fume and the water reducing agent, adding water, and uniformly stirring to obtain the cement-based material. The method provided by the invention can effectively relieve the problem of performance degradation caused by concrete structure cracking caused by the pelelith aggregate.

The invention discloses a threshold setting method and system under different working conditions based on fusion of MVMD noise reduction, KNN classification and an adaptive threshold. Acquiring n groups of vibration signal data of the equipment at different moments; performing MVMD decomposition to obtain a plurality of basic mode components IMF; calculating a correlation coefficient of each IMF and an original signal; the adding module is used for adding the basic mode components IMF of which the correlation coefficient is greater than a correlation coefficient threshold value to obtain a reconstruction signal; calculating a speed effective value and kurtosis of each group of reconstructed signals; setting KNN classification parameters, and selecting a parameter k with the maximum accuracy rate; threshold values under different working conditions are calculated; and determining a final threshold according to the standard value. A KNN classification method is adopted to separate characteristic values of different operation conditions of equipment, and then an adaptive threshold method is adopted to calculate thresholds of different conditions. The defect that a traditional method cannot set threshold values according to working conditions to cause pre-alarm is overcome.

The present invention discloses a method and system for setting thresholds under different working conditions based on the fusion of MVMD noise reduction, KNN classification and adaptive threshold. Obtain n groups of vibration signal data of the equipment at different times; perform MVMD decomposition to obtain multiple basic modal components IMF; calculate the correlation coefficient between each IMF and the original signal; add the basic modal components IMF whose correlation coefficient is greater than the correlation coefficient threshold to obtain a reconstructed signal; calculate the effective value and kurtosis of the velocity of each group of reconstructed signals; set the KNN classification parameters, select the parameter k with the highest accuracy; calculate the thresholds under different working conditions; determine the final threshold according to the standard value. The KNN classification method is used to separate the characteristic values ​​of different operating conditions of the equipment, and then the adaptive threshold method is used to calculate the thresholds of different working conditions. Different from the defect of the traditional method that cannot set the threshold according to the working condition and cause a pre-alarm.

The application discloses a method for alleviating Non-IID influence based on explainable federated learning. The application mainly introduces a validation set explainable mechanism to depict the explainable results of the validation set samples in the center server based on the influence of local client update on the learning ability of each category of the aggregated model, and evaluates the explainable results of each category by using a structural similarity index (SSIM), so as to deduce the data imbalance clients. Then, the parameters of the data imbalance clients are adjusted, the gradient distance between the model of the data imbalance client and the parameter of the global aggregated model in the last round is minimized, and the parameters of the model of the data imbalance client are corrected through the convergence of the gradient distance, so that the negative influence caused by the data imbalance is weakened.

The invention relates to the field of control, and discloses a switching system control system and method based on a TOD protocol and quantization constraint, and the system comprises an actuator, a switching system, a sensor, a communication network with the TOD protocol and a quantizer, an observer, and a controller. The sensor is provided with a plurality of nodes, and each node is responsible for processing measurement output of the switching system; the communication network with the TOD protocol and the quantizer is used for scheduling sensor nodes and quantizing measurement output; the observer is used for estimating the system state according to the quantized measurement output and sending a corresponding state estimation value to the controller according to the switching signal; and the controller is used for generating control input according to the state estimation value, controlling the switching system and ensuring that the system has no large buffeting amplitude. According to the system and the method disclosed by the invention, the problem of system performance reduction caused by data transmission congestion in a communication network can be solved.

The invention discloses a biaxially stretched polyolefin film and a preparation method thereof, and relates to the technical field of polyolefin films. The biaxially stretched polyolefin film is prepared from the following components by mass percentage: 1-8wt% of a cycloolefin polymer with a glass transition temperature of 120-170°C, 91-97wt% of a partially crystallized propylene homopolymer with a crystal melting temperature of 150-170°C, and 0-4wt% of an additive. The cycloolefin polymer content of 1-8wt% can significantly improve the heat resistance, dielectric properties and dimensional stability of the film material without excessively affecting the processing performance and cost. The 91-97wt% high-purity propylene homopolymer with a narrow molecular weight distribution ensures the excellent mechanical and electrical properties of the film material. The prepared polyolefin film has higher crystal melting point and glass transition temperature while maintaining good processing performance and economy, and has higher high-temperature dielectric strength, more excellent heat resistance and dimensional stability.

The present invention provides a semi-supervised medical image classification method based on a secure contrast self-integration framework, comprising the following steps: first, a batch of data consisting of labeled and unlabeled data is selected to add data perturbation to it, and the process is repeated twice to obtain two groups of data, which are respectively input into the student and teacher networks; then, a weight function is designed and the weight function is updated using supervised loss to automatically assign weights to each unlabeled data; a reliable perceived consistency loss is established by combining the weight parameters and the probability output of the two networks; further, a reliable perceived contrast loss is established by combining the weight parameters and using a projection network to obtain a normalized low-dimensional representation of the output of the two networks; finally, all loss functions are weighted and summed to form a final loss function, and network parameters and weight function parameters are updated alternately. The method proposed by the present invention can simultaneously learn reliable information at the data level and the data structure level, thereby improving the robustness and generalization of the model.

The invention provides a design method of a high-speed switch valve cooling flow channel considering both energy consumption and heat dissipation performance, which is characterized in that the cooling flow channel is a double-outlet series-parallel spiral cooling flow channel arranged in a switch valve shell, and when a switch valve works, a cooling medium is introduced into the cooling flow channel to enable the cooling medium to perform forced convection heat exchange with a switch valve component; heat generated when the switch valve works is effectively dissipated so as to reduce the temperature rise of the high-speed switch valve; according to the design method, by considering corresponding high-speed switch valve energy consumption and heat dissipation performance indexes under different flow channel parameter combinations, the matching of the flow channel screw pitch, the cross section transverse-longitudinal ratio and the inlet flow speed is designed; polynomial fitting is carried out according to the corresponding relation between different flow channel parameters and energy consumption and heat dissipation performance; solving the multi-objective optimization model by adopting an intelligent optimization algorithm to obtain a Pareto frontier solution set; and finally, carrying out comprehensive evaluation on the Paroto frontier solution set to obtain an optimal flow channel parameter. According to the method, the optimization efficiency is greatly improved, and the design period is shortened.

The application discloses an anti-poisoning attack cooperative security defense method and system for Internet of Vehicles, and relates to the technical field of Internet of Vehicles security. The method comprises the following steps: an initial model is issued by a cloud end to all vehicles at a vehicle end, vehicle end synchronously acquires in-vehicle network data and off-vehicle communication data, and performs preprocessing; the vehicle end trains and fine-tunes the received initial model with the minimum reconstruction error or feature space compactness loss as the target, and encrypts and updates the fine-tuned model based on privacy protection processing; the cloud end performs security screening on the encrypted and updated model parameters based on a multi-stage robust aggregation mechanism and a dynamic trust management model to obtain a global model; and the cloud end issues the global model and corresponding data to the vehicle end for real-time monitoring and feedback. The application can effectively resist malicious vehicle end attacks while guaranteeing vehicle data privacy and system real-time performance, and has strong cross-domain generalization capability.

The invention discloses a heterogeneous perception task scheduling method and system based on personalized federal reinforcement learning, and the method comprises the steps: constructing a task scheduling environment with a scene boundary based on load data, and constructing a local scheduling model based on the task scheduling environment through employing a dual-reviewer near-end strategy optimization algorithm; according to the method, public reviewer network model parameters in a constructed local scheduling model are aggregated, a personalized model is generated based on multi-head attention weights, and value estimation of local reviews and public reviews is fused by introducing Dual-Critic PPO, so that the problem of performance reduction of a global model in a heterogeneous environment is effectively relieved; and meanwhile, the stability of advantage estimation and the reliability of strategy updating are improved. According to the method, a server side carries out personalized aggregation on public reviewer parameters through a multi-head attention mechanism, so that different clients can obtain public models matched with environment characteristics of the clients instead of depending on a general average model, and thus faster convergence and stronger personalized adaptation are realized.

The present application relates to the field of handle error correction, and more particularly to a dynamic correction method and device for gamepad joystick offset error. The method comprises the following steps: detecting the player's no-input operation state, continuously collecting the original coordinate input value of the gamepad, and performing no-operation drift analysis to construct a static joystick dead zone; detecting the real-time operation signal stream of the player's handle, predicting the output behavior of the character in the scene, and generating game operation behavior data; based on the game operation behavior data, performing dynamic response deviation calculation, and constructing a handle drift vector field according to the static joystick dead zone; performing player operation intention analysis and unintended signal elimination processing on the real-time operation signal stream to obtain effective operation signals; performing reverse offset compensation calculation on the effective operation signals according to the handle drift vector field, and performing real-time joystick error correction. The present application corrects the signal recognition error of the gamepad, improves the response efficiency and accuracy of the game operation.

The invention provides an electrolyte, a preparation method thereof and a battery. The electrolyte comprises an electrolyte, an organic solvent and an additive, the additive comprises a functional additive, the functional additive meets the structural general formula (I): # imgabs 0 #, at least one of R3 and R4 in the structural general formula (I) meets the structural general formula (II): # imgabs 1 #, n is a natural number, and n is a natural number. R1, R2, R3, R4 and R5 are independently selected from any one of hydrogen, halogen, a cyano group, an isocyanate group, an alkyl group, an alkoxy group, a carboxylic ester group, an alkenyl group, an alkynyl group, an aryl group and a benzyl group. The electrolyte has stable electrochemical performance in a relatively wide temperature range, and the problem that the performance of the battery is reduced at an extreme temperature is relieved.

The application discloses a multilingual neural machine translation system using language family information, which is constructed based on a Transformer basic model, and a language family information module is added after each self-attention mechanism module and a feedforward neural network module in the Transformer basic model; the language family information module inputs language family information and the output of a preceding module, and outputs a vector after the language family information is fused. The application further provides a multilingual neural machine translation method using language family information. The multilingual neural machine translation system using language family information provided by the application alleviates the performance decline problem of large-scale multilingual machine translation on high-resource languages, and further improves the translation performance on low-resource languages.

Embodiments of the present disclosure provide an artificial intelligence chip, an artificial intelligence processor core and an electronic device. The artificial intelligence chip comprises a plurality of artificial intelligence processor cores and at least one communication chain, wherein the at least one communication chain comprises a first communication chain, the first communication chain sequentially connects the plurality of artificial intelligence processor cores in communication respectively, and comprises a plurality of first bidirectional communication units corresponding to the plurality of artificial intelligence processor cores respectively, each first bidirectional communication unit is connected in communication with a corresponding artificial intelligence processor core, and is configured to select a direction of data transmission of the corresponding artificial intelligence processor core through the first communication chain according to a received first control signal. The artificial intelligence chip can effectively solve the problem of system performance degradation caused by fixed direction communication bottleneck, and improve the overall efficiency of the artificial intelligence processor in processing complex workloads.