128 results about "Training period" patented technology

The invention discloses a reinforcement learning model training method and device based on dynamic data selection and electronic equipment, and the method comprises the steps: carrying out the screening from an original mathematical problem data pool according to preset data quality, diversity and difficulty evaluation indexes, and obtaining a candidate training set; calculating a data validity score of each training sample in the candidate training set; difficulty filtering is carried out on the candidate training set, and an adaptive training course is constructed; further filtering the candidate training set according to the solution reasoning process and the reward score of the candidate training set; dynamically selecting a data subset for the current training period by setting a temperature parameter attenuated along with the training process and a probability threshold based on the further filtered candidate training set and the data validity score; the selected data subsets are put into a reinforcement learning model for training, and the reinforcement learning model is trained in an optimized mode according to the correctness reward and the format reward. The model training period can be remarkably shortened, and the performance equivalent to or even better than that of full data training is achieved.

The invention discloses a medical auxiliary diagnosis method and electronic equipment based on quantum federated learning in the technical field of smart medical, and the method comprises the steps: carrying out the medical auxiliary diagnosis based on a pre-trained quantum federated learning model, and the obtaining process of the quantum federated learning model comprises the steps: constructing the quantum federated learning model, comprising a local client and a central server. Repeatedly executing the following steps until a preset training period is reached: in the current training period, training the local client according to the training sample set, and calculating the accuracy of the local client according to the test sample set; the local client sends the accuracy and the final model gradient to the central server after the current training period; the central server updates global model parameters according to the accuracy and the model gradient, and returns the global model parameters to the local client; and the local client updates the local model parameters according to the global model parameters. According to the invention, the accuracy and efficiency of medical auxiliary diagnosis can be improved.

The invention belongs to the technical field of rehabilitation training, and particularly relates to an active stroke rehabilitation auxiliary training system which generates a personalized training base line through clinical evaluation data of a patient and guides the patient to cooperatively execute training actions through rehabilitation equipment according to the training base line. Training execution efficiency scores, patient active motion features and equipment auxiliary performance features are analyzed in real time, and rehabilitation equipment auxiliary strength and training task difficulty are dynamically adjusted in the training duration according to the double-feature real-time interaction state and the synchronous training execution efficiency scores; finally, the rehabilitation ability state of the patient is updated according to the multi-dimensional time sequence characteristics of the whole single training process, and a baseline correction instruction of the next training period is output based on the updated rehabilitation ability state, so that the limitation of singly paying attention to patient actions in the past is broken, and the boundary of patient independent effort and equipment assistance in the rehabilitation process is clearly distinguished; and the dynamic adaptability and accuracy of stroke rehabilitation training are effectively improved.

The invention provides a concentration training method and device based on visual tracking, and the method comprises the steps: collecting the real-time eye movement data of a user, including the staring point coordinate and the eye movement speed, and initializing the target trajectory of a training task; calculating a concentration deviation between the current gazing point and the target trajectory, and dynamically adjusting the target trajectory through nonlinear mapping based on the deviation to generate a dynamic trajectory; constructing a multi-feature fusion distraction scoring model based on the concentration deviation and the change trend of the concentration deviation in combination with the geometric features of the trajectory, and detecting and marking distraction events; when a distraction event is detected, self-adaptive visual correction stimulation is executed at the event position immediately, and a user is guided to return to a task; and after the training period is finished, dynamically adjusting the task difficulty coefficient of the next period according to the frequency and the average deviation of the distraction events in the period.

The invention provides a method and device for determining a training plan, electronic equipment and a storage medium, the method is applied to the technical field of exercise training, and the method comprises the following steps: in response to a training plan generation request for a target competition, obtaining a current physical fitness parameter of a target user and a residual training duration of the target competition, the target physical fitness parameters of the target competition are obtained; determining a total training load according to the current physical fitness parameter, the target physical fitness parameter and the residual training duration; based on the total training load and the residual training duration, period training loads of a plurality of training periods in the residual training duration are determined; and generating a training plan of each training cycle based on the cycle training loads of the plurality of training cycles. According to the method, when the training plan is made for the user, the generated training plan has pertinence and meets the requirements of the target competition and the training requirements of the user, the user is prevented from being over-trained, and therefore the user can obtain the best training effect in the training process.

A computer implemented method of access control for a user device having at least one component for determining behaviors of the user. The method including defining a training period during which access to the device is determined based on a credential-based authentication scheme wherein each access determination is used to generate an item of training data and training a machine learning classifier based on the training data such that the classifier is operable to classify user behavior as compliant or non-compliant. The method further including, in response to a determination that a behavior subsequent to the training period is classified as non-compliant, requesting a credential-based authentication of the user and permitting access to the device in response to the credential-based authentication, wherein permitting access to the device further includes constructively training the classifier based on the subsequent behavior as a compliant behavior.

The invention discloses an intelligent data analysis system and method based on source-network-load-storage integration, and relates to the technical field of electric power analysis, the method comprises the steps: carrying out the statistics of the historical average power consumption of a certain time period, and determining a characteristic time period; a training period is set, the electricity consumption of the load end at the sampling time point is collected, and the electricity consumption fluctuation rate of the load end in the characteristic time period is calculated; drawing an electricity consumption fluctuation graph of the load end in the characteristic time period in the training period, and determining a target load end; generating an energy storage end group, collecting operation parameters of the energy storage end group, calculating an operation score of the energy storage end group, and drawing a feature operation score graph of an analysis period corresponding to the energy storage end group; and acquiring a real-time operation score of the energy storage end group, calculating a real-time analysis period of the energy storage end group, judging an abnormal analysis state of the energy storage end group, and ensuring efficient and stable operation of the source-grid-load-storage integrated power system.

The invention discloses an image content recognition system and method based on artificial intelligence, and relates to the technical field of data analysis, and the method comprises the steps: carrying out the recognition through collecting the identity and image information of a person, recording a failure reason, calculating the occurrence frequency of the failure reason, and determining a failure level; historical identification failure records are collected, the occurrence frequency of abnormal data is calculated, and target data are determined; setting a training period, dividing a day into a plurality of characteristic time periods, calculating failure cause frequency in each time period, and determining a target time period; counting the number of identification records in a target time period of a certain day, calculating a failure frequency and determining an optimization period; and calculating the real-time failure frequency in the optimization period. If so, reminding a worker; if yes, the optimization period is adjusted, and therefore the utilization efficiency of system resources is maximized while the recognition accuracy is improved.

The invention provides a model training method, mobile equipment, electronic equipment and a medium, and the model training method comprises the steps: executing a test task through a task execution model, and obtaining test data corresponding to a target time period in the process of executing the test task; under the condition that the test result corresponding to the test data does not meet the preset requirement, correcting the test data to obtain corrected data; and training the task execution model by using the correction data to obtain a trained task execution model. According to the technical scheme, the whole training period of the model can be shortened, and the training efficiency is improved.

The invention relates to the technical field of unmanned system multi-agent control, in particular to a wheel type mobile mechanical arm coordination control method, system and device. Comprising the following steps that a wheeled mobile mechanical arm model is established, and a target reachability area is solved; the operability index of the wheeled mobile mechanical arm is calculated; calculating the maximum linear velocity and angular velocity of the wheel type mobile mechanical arm to obtain a velocity ellipsoid model of the wheel type mobile mechanical arm; designing a multi-agent combined state space, and designing a multi-agent combined action space; constructing a control algorithm flow; setting a task scene of a coordination control algorithm of the wheeled mobile manipulator, setting training hyper-parameters, and designing task evaluation indexes; and carrying out training setting, and recording the success rate of the task and the validity of the process when the training period is finished. According to the method, the problem that the model structure is complex and depends on kinematics equation solving or objective function optimization under the high-dimensional condition is solved. According to the invention, the technical effect of reducing complexity calculation is achieved.

The invention relates to the technical field of model training sample processing, in particular to a sample distribution adaptive adjustment method and system based on model training feedback driving, and the method comprises the following steps: S1, collecting feedback indexes in a training process in real time in a model training period, and obtaining a feedback index set; s2, weak item categories are analyzed and recognized based on the feedback index set, and an optimization instruction is generated; s3, dynamically adjusting the sampling weight to form an updated sampling strategy; s4, samples are collected according to a sampling strategy, and weak item category samples are supplemented; s5, putting the optimized sample set into the next round of training, and circulating the steps S1 to S4 until the training indexes of the learning effects of all key categories meet the termination condition; and acquiring a sampling strategy and a sample set. According to the method, real-time, refined and automatic sampling adjustment can be realized by taking the actual performance of model training as a core feedback index, so that the training effect and generalization ability of the model in key areas such as weak-term categories and long-tail categories are improved.

The invention provides a language model training method and device, electronic equipment and a storage medium. The method comprises the following steps: training an initial language model to obtain a weight matrix, and decomposing the weight matrix to obtain a basic feature matrix and an initial task adaptation matrix; updating the initial task adaptation matrix to obtain an updated task adaptation matrix; determining the current sensitivity of each singular vector direction in the update task adaptation matrix of the current training period; determining a sensitive direction from a plurality of singular vector directions according to the current sensitivity, and performing regularization processing on a scaling coefficient associated with the sensitive direction to obtain a regularization loss function; in response to determining that the change sensitivity of the singular vector direction in the continuous preset number of training periods is in an attenuation state, performing parameter resetting on the zoom coefficient to obtain an updated zoom coefficient; and carrying out merging processing on the basic feature matrix and the update task adaptation matrix, and taking the update language model as a target language model.

The invention relates to the technical field of medical training, and discloses a sputum suction operation training method and system based on a virtual reality technology. The method comprises the steps that historical operation records of a trainee are collected, key operation nodes are extracted, and an airway state model of a virtual patient is constructed based on the nodes; secondly, setting a virtual sputum suction equipment model, performing a multi-working-condition simulation test on the virtual sputum suction equipment model to obtain operation response characteristics, and mapping the equipment model into the virtual airway state model; then calculating a secretion distribution parameter of a virtual airway in a training period and a theoretical clearance load of virtual equipment, and combining operation response characteristics to obtain an operation intensity parameter; generating an operation triggering instruction sequence based on the operation intensity parameter, and performing staged operation control on the virtual equipment according to the sequence; and finally, carrying out quantitative evaluation on the operation effect of the virtual equipment in the virtual airway. According to the method, a targeted virtual training scene can be constructed, a real operation response is restored, and accurate tracking and effect quantitative evaluation of the operation process are realized.

The invention relates to the field of basketball teaching, and discloses a basketball teaching feedback system and method based on action recognition, and the method comprises the steps: employing a multi-sensor cooperative collection mechanism to carry out the partition sampling of body key points and a basketball moving track in a fixed-point shooting training process of a student, and obtaining an original signal set for depicting detailed actions; the original signal set is mapped into a multi-stage time sequence model, and modeling is conducted on the nonlinear coupling relation between the wrist rotation amplitude and the elbow joint stretching angle; based on the multi-dimensional feature tensor, constructing a multi-scale sliding window to dynamically slice the angle track at the moment of hand release; generating a grading heat map according to the influence degree of the key abnormal segment on the shooting precision, and respectively endowing different priority weights to the hand-out angle stability, the elbow joint control force and the follow-up coherence; and based on the feedback priority queue, carrying out dynamic evolution deduction on the chain abnormity formed in the training period. The method has the advantage of improving the pertinence of teaching.

The invention relates to the technical field of medical rehabilitation equipment, in particular to a surgical postoperative rehabilitation nursing auxiliary device and method based on the Internet of Things, and the device comprises a bearing part, a driving part, a base, a monitoring part and a remote interaction part, the driving part comprises a first driving assembly for providing relative resistance for a using object and a second driving assembly for driving the thigh and the shank to move relatively. The method comprises the steps that in a single training period, the passive muscle activation degree and the active muscle activation degree are determined to determine the rehabilitation degree of the muscle function of a using object; according to the limb movement track and the muscle strength balance condition of the feature movement part, the movement control ability of the using object is determined, and a rehabilitation training scheme of the using object is adjusted according to the rehabilitation degree or the movement control ability. According to the invention, comprehensive monitoring and remote transmission of the rehabilitation process of the use object are realized, and the rehabilitation process can be accurately controlled by medical personnel.

The invention provides a model training correction method and related equipment, and relates to the technical field of artificial intelligence, and the method comprises the steps: collecting the historical gradient information of the network parameters of a plurality of network layers in real time during each iteration updating in the iterative training process of a large model; determining a correction strategy according to the historical gradient information; adjusting the correction amplitude of the correction strategy according to the network layer of the large model, and determining the correction amplitude of the current correction; and correcting the current gradient information of each network layer of the large model according to the correction strategy and the correction amplitude of the current correction to obtain corrected gradient information. By monitoring and analyzing historical gradient information in the iterative training process, the gradient updating direction and amplitude are adaptively corrected, the oscillation and instability phenomena in the later training period are reduced, and the training stability and convergence speed are improved.

The invention relates to the field of mode recognition, and discloses a traditional physical health intervention system and method combined with multi-mode data, and the traditional physical health intervention method combined with the multi-mode data comprises the steps: continuously collecting physiological monitoring signals, action execution records and training scene parameters; assigning differentiated attention weights to various monitoring segments; performing cross-channel feature comparison on the real-time monitoring segment, and extracting a comprehensive training state feature group of the body load bearing capacity and the action control state of the training participant; based on the comprehensive training state feature group, identifying an action risk fragment influencing the training safety; performing reordering according to the concentration degree of the action risk fragment triggering conditions, and constructing a training risk labeling set of the potential risk change range and degree; and through training the risk labeling set, determining a key period with a relatively high risk accumulation speed in a specific training period. The device has the advantage that the safety of traditional physical training is improved.

The invention provides a knowledge distillation-based model compression method and system, and the method comprises the steps: inheriting a soft label of a maturely trained teacher model, defining a distillation loss function of a student model through the soft label, enabling the student model to be compressed, generating a new sample extension training set through random sample extraction and mixing, and achieving the compression of the student model. The problem of insufficient data is relieved, meanwhile, the robustness of the model to noise and distribution offset is improved, through dynamic weight adjustment, a student model is made to quickly fit teacher model knowledge, autonomous optimization is conducted in the later training period, and over-fitting of soft labels is avoided. On a test set in which the proportion of unlabeled data is 30%, on the basis that the accuracy of using a real label is 78.3%, the accuracy is improved by 11.2% by using a pseudo target, deployment is performed by using a maturely trained student model, middle layer feature alignment is omitted, the occupation of a GPU video memory is reduced from 3.2 GB to 1.8 GB, and the deployment requirement of edge equipment is met.

The invention discloses an intelligent diopter training system and method based on a multi-modal physiological feedback and reinforcement learning algorithm. The intelligent diopter training system comprises a multi-modal physiological data acquisition module, a central processing PC, a user terminal and a cloud data platform. Visual behavior data, ciliary muscle state data and autonomic nerve reaction data of a user are collected in real time through a multi-modal physiological data collection module, multi-modal physiological data fusion and feature extraction are carried out, and a refraction state evaluation model is used for calculating a comprehensive score of a refraction state grade, visual fatigue and adjustment sensitivity of the current user. A personalized training parameter set of the next training period is dynamically decided and generated based on a reinforcement learning algorithm, and the user terminal presents visual training stimulation according to the current personalized training parameter set. The cloud data platform stores the data and generates a visual training effect trend chart, a diopter change prediction report and personalized training suggestions to complete training scheme personalization, training means intellectualization and feedback evaluation comprehensiveness.

The invention discloses a training method of a credit scoring model, a credit scoring method and a related device. When the method provided by the embodiment of the invention is executed, a sample training set containing a plurality of credit score labels is obtained, each sample not only has the credit score label, but also comprises a plurality of sample data marked with risk modes, the risk mode labels indicate the monotonic relationship between the sample data and credit scores, and the credit score labels indicate the monotonic relationship between the sample data and the credit scores. Comprising monotone increasing, monotone decreasing or non-monotone relation, and the sample data covers credit investigation information of the sample user. On this basis, a back propagation algorithm is used to train a full-connection neural network (i.e., an initial scoring model) constructed based on a combined activation function, thereby obtaining an optimized credit scoring model. According to the method, the sample risk mode label is introduced into the training set, and the credit scoring model is constructed by using the combined activation function, so that the definite definition and capture of the monotonicity relation are realized, and the model is ensured to follow the monotonicity constraint during the training period.

The invention relates to the technical field of federated learning in an Internet of Vehicles scene, in particular to a federated learning method suitable for a high dynamic environment of a vehicle. According to the method, adaptive vehicle federation learning is provided, so that a local training period adapts to the vehicle capability, the scheduling flexibility is improved, the influence of insufficient local period on model updating is relieved, the training efficiency is improved, and the communication competition is reduced. Meanwhile, a federal learning scheduler based on an actor-commentator structure is designed to realize a dual-time scale optimization method, so that the training time is minimized and the model performance is maximized. According to the method, the vehicle federated learning performance reduction caused by the dynamic complex vehicle edge network is effectively relieved, and the training efficiency of the vehicle federated learning is improved.

The invention discloses an exercise training recommendation method and system for disability risk crowds. According to the method, a multi-level portrait system of a patient is constructed, multi-dimensional data such as disease types, clinical features, demographic information and motion behavior features of the patient are fused, role portraits are generated, and static features are obtained. After the first exercise training scheme is pushed based on the static features, structural data, text type data and image type data of the patient in the training period are collected, a subsequent exercise training scheme is generated through a collaborative filtering model, and iterative optimization is conducted on the dynamic features of the patient through a recurrent neural network model. The exercise training scheme can be dynamically adjusted according to the individual features and training feedback of the patient, and the compliance and the training effect of the patient are improved.

The invention discloses a mechanical arm teleoperation training method based on force feedback, and the method comprises the steps: constructing a hot chamber virtual environment containing a mechanical arm model and an operation object through a scene simulation platform Unity, and configuring a joint position according to a real mechanical arm DH parameter; the dynamics simulation platform AppeliaSim builds a corresponding dynamics model, and an inverse kinematics solution and collision detection function is configured; integrating a force feedback device, and constructing a working space mapping model to convert hand motion into an expected pose of the tail end of the mechanical arm; the double platforms synchronize data through a dynamic link library, and the object control and force sense calculation module calculates collision force and bolt assembling and disassembling torque and transmits the collision force and the bolt assembling and disassembling torque to force feedback equipment. The method does not need to depend on a real mechanical arm, safety risks and high cost are avoided, training repeatability and pertinence are improved, immersion and operation precision are enhanced through force sense and visual sense synchronous feedback, the training period is shortened, and the method is suitable for fine operation training of the mechanical arm in a dangerous environment.

The utility model discloses a scapulohumeral periarthritis rehabilitation training device which comprises a main body frame, a handle, a mooring rope, a chain, a two-way chain wheel assembly and a wheel carrier, the wheel carrier is arranged below the top of the main body frame, the two-way chain wheel assembly is arranged in the wheel carrier, and the chain penetrates through the two-way chain wheel assembly and can drive a chain wheel to rotate in two directions. The two ends of the chain are respectively connected with a cable, the tail ends of the two cables are connected with handles, and a bidirectional ratchet wheel structure is arranged in the bidirectional chain wheel assembly. Through the arrangement of the bidirectional chain wheel assembly with the bidirectional ratchet wheel structure, the tension meter, the graduated scale, the damping fin and other structures, fixation at any height within the design range and fixation training of gradually increasing the height can be achieved in the training process, and the handle cannot be loosened in the training process; the pulling damping is adjustable, and the rehabilitation training device can adapt to rehabilitation training of patients with different strengths; the tension and height in the training process can be observed in real time, and the effects of different training periods can be effectively mastered.

The invention discloses a return anxiety intervention method and system based on VR scene self-adaption and data feedback, and relates to the technical field of return anxiety intervention methods.The return anxiety intervention method comprises the steps that occupational attributes, disease stages and psychological state evaluation data of a user are collected, and physiological baseline information obtained by wearable equipment is combined; constructing an initial user portrait containing individual features and function states; matching a preset intervention path template according to the initial user portrait, and generating a personalized VR training scheme with a module sequence, a scene difficulty level and a training period; executing the personalized training scheme in a virtual reality environment, carrying out commuting simulation, task operation and social interaction training through an immersive office scene, and synchronously collecting multi-modal physiological response data of a user in a training process; the anxiety level of the user is dynamically evaluated based on physiological response data in the training process, and when it is detected that the anxiety state exceeds a preset tolerance range, the stimulation intensity of the current scene is automatically reduced, and a relaxation bootstrap program is started.

The invention discloses a virtual sensory integration training method based on force feedback, and the method comprises the steps: constructing a hot chamber virtual environment containing a mechanical arm model and an operation object through a scene simulation platform, and configuring a joint position according to a real mechanical arm parameter; the dynamics simulation platform builds a corresponding dynamics model and configures inverse kinematics solution and collision detection functions; force feedback equipment is integrated, and a working space mapping model is constructed to convert hand motion into the expected pose of the tail end of the mechanical arm; the double platforms synchronize data through a dynamic link library, and a force sense calculation module calculates collision force and bolt assembling and disassembling torque and transmits the collision force and the bolt assembling and disassembling torque to force feedback equipment. The method does not need to depend on a real mechanical arm, safety risks and high cost are avoided, training repeatability and pertinence are improved, immersion and operation precision are enhanced through force sense and visual sense synchronous feedback, the training period is shortened, and the method is suitable for fine operation training of the mechanical arm in a dangerous environment.

The application provides a rule cultivation quality full life cycle monitoring method and system, which comprises the following steps: in a training period, collecting operation logs of rule cultivation personnel in multiple clinical information systems, and reconstructing the operation logs into behavior trajectory sequences reflecting clinical core capabilities; evaluating the quality deviation degree of the rule cultivation personnel in the clinical case processing process based on the behavior trajectory sequences; obtaining the skill mastering vector of the rule cultivation personnel in each clinical core capability based on the quality deviation degrees of multiple clinical operations in a historical period and a preset capability dimension mapping rule; generating the capability evolution characteristics of the rule cultivation personnel in the training period according to the change law of the skill mastering vector over time; and generating a multi-dimensional evaluation result including skill maturity, risk deviation degree, diagnosis and treatment logic consistency and clinical efficiency factors based on the capability evolution characteristics and the quality deviation degree. The application improves the timeliness and accuracy of rule cultivation quality evaluation.

The application belongs to the field of police dog training system, and provides a police dog search smell ability automatic training system, which comprises a base and a support frame, a function frame fixedly arranged on the base, an odor generating mechanism arranged on the function frame, a driving mechanism arranged in the function frame, a connecting mechanism arranged between the function frame and the support frame, and a shaping mechanism slidingly arranged on the support frame. Through conditional automatic reward to the police dog, the key elements in the police dog search smell training are solved, the purpose of police dog search behavior training under the condition of no monitoring is realized, the positive reinforcement method is adopted to make the police dog operation more active and lively, meanwhile, unmanned autonomous training can be realized, passive reaction of the police dog is prevented, the training and the police dog life scene are integrated, and the training efficiency can be greatly improved, the training period can be shortened, and the labor cost and expense can be saved.

The invention discloses a fitness intensity self-adaptive adjusting method and system based on physiological data feedback, and relates to the technical field of medical rehabilitation. The method comprises the following steps: acquiring historical physiological limit data of a target rehabilitation user and an initial physiological state baseline of a current training period; calculating a dynamic safety intensity interval of the user individual; collecting physiological feedback data in real time in training, inputting the physiological feedback data into the stimulation efficiency evaluation network, and outputting an actual physiological load value and an effective stimulation coefficient of current training; calculating the recommendation intensity of the next training unit in the safety interval according to the effective stimulation coefficient and the rehabilitation progress function; and executing the recommendation intensity, updating the evaluation value based on a new round of physiological feedback data, and carrying out loop iteration until a training cycle target is reached. According to the invention, through dynamic safety boundary and intelligent closed-loop regulation and control, personalized and adaptive optimization of medical rehabilitation training intensity under safety guarantee is realized, and the safety and efficiency of training are improved.

The application provides a trigger type wearable ultrasonic device and method, and relates to the technical field of medical imaging and rehabilitation engineering. The device comprises: an ultrasonic imaging module for generating an ultrasonic image of a target part; a training intensity monitoring module for monitoring rehabilitation training actions of a user in real time and generating a training intensity signal; a central processing and control module for generating an action type and a training intensity parameter according to the training intensity signal, generating an imaging trigger instruction according to a trigger condition, and synchronously and associatively storing the collected ultrasonic image and the training intensity parameter and the action type at the trigger time. Thus, the ultrasonic imaging is intelligently triggered through the training intensity signal, the system only works in an effective training period, thereby significantly reducing the power consumption of the system and prolonging the endurance time; and the multimodal synchronization and association of the ultrasonic image data and the training intensity parameter are realized, thereby providing an objective and fused data basis for the quantitative evaluation of rehabilitation effects and the accurate formulation of rehabilitation plans.