7results about How to "Improve self-learning ability" patented technology

The invention discloses an interactive load prediction method based on fusion of prior knowledge and data driving, and the method comprises the steps: reconstructing a load sequence into a trend load sequence and a disturbance load sequence through modal decomposition, and constructing a prior knowledge base of tag-feature weight according to SHAP interpretability analysis; and through a prior knowledge base and feature contribution analysis, channel-level weight learning is carried out by using an automatic correlation determination module, and key input features are obtained. And inputting the reconstructed load sequence and the key input features into an improved Transform model, mapping a priori knowledge base into a priori knowledge matrix, embedding the priori knowledge matrix into the model, and generating a load prediction result. And carrying out interpretability analysis on a prediction result, and feeding back and updating an analysis result to a priori knowledge base to realize dynamic collaborative updating of prediction and the knowledge base. According to the invention, organic combination of expert knowledge and a data-driven model is realized, and the precision and stability of load prediction are improved.

This invention discloses a closed-loop control method and system for prevention, intervention, and risk feedback in cerebral hemorrhage, relating to the field of cerebral hemorrhage risk intervention technology. It includes the following steps: S1, real-time acquisition of multi-source cerebrovascular monitoring data, data preprocessing, and identification of high-pressure peak events; S2, quantification of key dynamic characteristics of each peak based on the multi-source cerebrovascular monitoring data corresponding to the high-pressure peak events, and assessment of high-pressure peak abnormalities; S3, fusion of multi-source cerebrovascular monitoring data and high-pressure peak abnormality assessment results, quantification of the risk probability of each peak event, identification of physiological fluctuations and risk events, and risk warning and response; S4, real-time monitoring of multi-source cerebrovascular monitoring data after risk warning, assessment of warning accuracy, identification of missed and false alarms, and threshold optimization. This solves the technical problems of existing technologies, such as difficulty in distinguishing between physiological and pathological peaks, susceptibility to misjudgments, and impact on the timeliness and individualized management of risk intervention.

The invention discloses an intelligent coffee machine operation state monitoring method and system based on reinforcement learning, and the method comprises the following steps: S1, collecting and preprocessing data, and forming an operation parameter sequence; s2, carrying out dimension reduction processing on the operation parameters, and constructing an operation state sequence; s3, performing multi-step prediction on the running state sequence through a DLinear model, and generating a state prediction sequence by adopting a linear predictor; s4, calculating an instant reward value of each regulation and control behavior according to the running state sequence and the state prediction sequence; s5, an A3C algorithm is adopted, and a regulation and control instruction is generated and executed according to the instant reward value; s6, establishing a state transition group and writing the state transition group into an empirical data set; and S7, updating the DLinear model and A3C algorithm parameters according to the empirical data set. According to the method, the Markov model, the principal component analysis, the DLinear model and the A3C algorithm are fused, and the method has the advantages of being high in adaptability, high in prediction precision and good in stability.

This invention relates to the field of scheduling technology, specifically to an intelligent MCU scheduling method based on adaptive edge learning. The method includes the following steps: acquiring the working status data, historical scheduling behavior data, and peripheral sensing signals of the MCU running node; constructing a task running scenario feature vector and an edge classification model based on the acquired task load parameters, peripheral call indicators, power consumption sampling data, and runtime; inputting the task running scenario feature vector into the edge classification model, performing an initial scheduling strategy matching operation, generating a candidate strategy set, and outputting scheduling strategy category labels; matching the corresponding scheduling control template according to the scheduling strategy category labels, and executing the MCU task scheduling process in conjunction with a preset task queue scheduling relationship; recording the context chain of abnormal events to construct a task abnormal behavior chain; and triggering the edge learning module to correct the scheduling strategy category labels based on the task abnormal behavior chain. This invention can improve resource utilization efficiency.

The invention provides a system for detecting mental diseases through face recognition based on an artificial intelligence technology, and the system comprises a collection unit which collects the original image data of a face through a plurality of cameras, and the original image data comprises static image data and dynamic image data; the static image data and the dynamic image data are respectively sent to a static analysis module and a dynamic analysis module; the static analysis module comprises a recognition unit and a first partition unit, the recognition unit obtains static key point data according to the static image data and recognizes facial expressions according to the static key point data, and the first partition unit divides a human face into a plurality of areas according to the static key point data; the dynamic analysis module is used for analyzing dynamic key point data obtained through dynamic image data under a preset condition, and the dynamic key point data comprises pupil action track parameters, mouth action track parameters and face jittering parameters.

The invention relates to a cloud firewall security protection method and system, a storage medium and an intelligent terminal, and relates to the field of cloud security technologies, and the method comprises the steps: obtaining target traffic; analyzing the target traffic to obtain a traffic application type and a traffic behavior activity; performing comparison based on the traffic application type and the threat signature; marking the target traffic as a potential threat when the matching is successful; when the matching is unsuccessful, searching a normal behavior from a normal database based on the traffic application type; performing matching based on the traffic behavior activity and the normal behavior; operation is not carried out when matching succeeds; when the matching is unsuccessful, performing deep behavior analysis to mark the target traffic as an advanced threat; and when the threats are marked as potential threats and advanced threats, the firewall adopts a protection action, and the method has the advantages that the DPI technology is combined with the firewall technology, so that the recognition rate of known threats is improved, and the detection and response capability of unknown or complex attack modes is enhanced.

The present invention relates to a method and system for adjusting the backlight brightness of a display screen based on Mini QLED technology. The method: divides a display image signal into a brightness signal, a chromaticity signal and a saturation signal, and generates image feature data; performs sub-pixel boundary detection to obtain multiple initial backlight sub-regions, and calculates regional brightness distribution data; performs regional clustering operations and regional merging to obtain multiple target backlight sub-regions and calculates backlight partition mapping data; performs nonlinear feature transformation, and calculates the backlight unit control data of the Mini LED backlight unit; uses a deep Q network to output the backlight compensation amount of each of the target backlight sub-regions; performs real-time feedback correction, calculates the deviation value between the actual display parameters and the target parameters of each target backlight sub-region, and outputs a backlight optimization control signal. The implementation of the present invention realizes the precise adjustment of the backlight brightness and improves the consistency of the display effect.