90results about How to "Avoid forgetting" patented technology

The invention discloses a federal forgetting learning method based on small sample knowledge distillation. The method comprises the following steps: initializing a global model by adopting a server and broadcasting the global model to each client; each client carries out local training on the received global model by using local data, and after the local training is completed, the global model is updated and sent back to the server; the server stores and aggregates the global models sent by all the clients, is used for updating the local global model of the server, and repeats the federal learning process from S1 to S3; a target client makes a forgetting request at a certain moment, federal learning is stopped, and at the moment, a server side uses a forgetting module to generate a forgetting model and takes the forgetting model as a student model; the server side finely adjusts and compares the language image pre-training model according to local small sample data, so that the language image pre-training model adapts to downstream data and serves as a teacher model; the server uses local small sample data to carry out recovery training through a knowledge distillation method, so that catastrophic forgetting caused by client data revocation is avoided, and the performance of a forgetting model is improved.

The invention relates to the technical field of artificial intelligence, and discloses a model updating method and device based on artificial intelligence, equipment and a medium, and the method comprises the steps: obtaining to-be-added data, and carrying out the data recall processing from an original model database according to the to-be-added data, so as to obtain an incremental data set containing the to-be-added data and the recall data; training the incremental data set by using a fine tuning training algorithm to update the original model; and performing preference optimization processing on the updated original model to obtain a target model preferred to the to-be-added data. Through incremental training and preference optimization processing, when the model is updated, forgetting of other old knowledge can be avoided, the old knowledge can be completely replaced for specific knowledge of the to-be-added data, the model does not need to be retrained, and training time and training resources can be effectively saved.

The embodiment of the invention discloses a model training method and device, electronic equipment, a medium and a program product, which can be applied to various scenes such as cloud technology, artificial intelligence, intelligent traffic, auxiliary driving and the like. According to the embodiment of the invention, a training sample set is obtained, wherein the training sample set comprises an initial sample text corresponding to an initial scene and a newly-added sample text corresponding to a newly-added scene; translating the initial sample text through a teacher model to obtain a teacher translation result; translating the initial sample text and the newly added sample text through an initial student model to obtain a student translation result; calculating a loss value by combining the teacher translation result and the student translation result; and on the basis of the loss value, updating model parameters of the initial student model to obtain a trained student model, so that the trained student model is used for translating the to-be-translated text corresponding to the initial scene and the newly-added scene. Therefore, according to the scheme, the multi-scene adaptive capacity of the model can be enhanced while the model training efficiency is improved.

The invention discloses a load prediction method and system based on error division multiple tasks, a medium and equipment, and the method comprises the steps: combining an error division task with a continuous learning mechanism, firstly calculating a sample relative error in a training process, classifying low-error data into a data set corresponding to a learned load feature mode, and carrying out the calculation of a data set corresponding to the learned load feature mode; the high-error data is delimited as a new task to be learned; measuring model parameter importance through a Fisher information matrix of EWC, introducing an L2 regularization item to constrain core parameter updating, and avoiding forgetting a learned old mode; and circularly executing task division and model training, so that a single model gradually masters various load characteristic modes in continuous learning. A division basis does not need to be set manually, a multi-mode learning process is simplified, and prediction accuracy and robustness in a complex load scene are improved.

The invention discloses a cloud desktop throughput prediction method for cold start and mode forgetting. The method comprises the steps of 1, learning aided training based on similarity attention federal; 2, throughput prediction for safe cold start in the initial stage of a new user is carried out; step 3, fluctuation-residual feature extraction based on a TDLinear network model is carried out; 4, updating the model based on the adaptive memory module and the EWC; and step 5, training forgotten throughput prediction in the long-term use stage of the user. Through a federated learning algorithm, on the premise of ensuring user data privacy, existing throughput historical use knowledge of other users is utilized to assist training. An old mode is continuously reviewed based on continuous incremental prediction, a memory module is designed to incrementally update a new mode, and the method has the advantages of avoiding forgetting in training and integrating two stages to realize full-life-cycle throughput prediction of the cloud desktop user.

The application provides a traffic classification method and system based on federal semi-supervised learning. The method comprises the following steps: constructing an unlabeled traffic dataset and a labeled traffic dataset; a center server decomposes a global model into supervised learning parameters and unsupervised learning parameters and initializes the parameters; the parameters and an auxiliary agent are sent to each client; each client performs unsupervised training based on the supervised learning parameters, the unsupervised learning parameters and the auxiliary agent by using a local unlabeled traffic dataset, uploads the difference between the unsupervised learning parameters to the center server; the center server aggregates and updates each unsupervised learning parameter; supervised training is performed by using a local labeled traffic dataset, and the difference between the supervised learning parameters and the difference between the unsupervised learning parameters are sent to each client; a new auxiliary agent is obtained based on nearest neighbor search, and the new auxiliary agent is sent to each client when a set sending condition is met; the two steps are iteratively executed until a stop condition is met.

The utility model provides a multifunctional tuning fork device for ear-nose-throat head and neck surgery, and relates to the technical field of medical instruments, the multifunctional tuning fork device mainly comprises a tuning fork, a connecting rod assembly and a hammer, the tuning fork is Y-shaped, the connecting rod assembly is fixedly arranged at the non-bifurcated end, the hammer is arranged at the tail end of the connecting rod assembly, and the connecting rod assembly is connected with the hammer. The connecting rod assembly comprises a connecting rod, a first handle and a second handle which are arranged in sequence, one end of the connecting rod is fixedly connected with the tuning fork, the opposite end of the connecting rod is detachably connected with one end of the first handle, the opposite end of the first handle is detachably connected with one end of the second handle, and a sound hammer is fixedly arranged at the opposite end of the second handle; according to the tuning fork device, when the tuning fork device is not used, the second handle and the first handle are connected together, so that forgetting to carry the tuning fork can be prevented, when the tuning fork device is used, the second handle and the first handle are detached, the tuning fork and the hammer are used in an impact mode, and a doctor can use the tuning fork device conveniently.

The invention discloses a bearing fault diagnosis method and device, a medium and equipment, and relates to the technical field of mechanical fault diagnosis. Comprising the following steps: constructing a plurality of data sets containing different bearing fault types, and obtaining a fault diagnosis model based on the data sets; and repeatedly carrying out trimming operation, data set merging operation and network training operation on the fault diagnosis model to obtain the fault diagnosis model for learning all bearing fault features. In order to balance the old knowledge retention capability and the new data learning capability of the model, old samples and new samples are selected from historical data to be combined, and partial parameters in the trained fault diagnosis model are frozen, so that the model avoids forgetting old task knowledge while updating the parameters of the model; meanwhile, in order to enhance the learning ability of new fault features, the model is pruned to overcome the negative influence of a fixed network model structure on new feature learning, and the learning ability of a new network model on new fault features is enhanced.

The invention provides a medicine taking management method and device, electronic equipment and a storage medium, the method comprises the steps that medicine taking reminding voice and a medicine taking reminding text of a target object in at least one set medicine taking time sent by a server side are obtained, and the medicine taking reminding voice of each medicine taking time is obtained according to the tone and the voice style of reference audio; the medicine taking reminding text of each medicine taking time is obtained by performing voice synthesis on the medicine taking reminding text of each medicine taking time, the medicine taking reminding text of each medicine taking time comprises target attribute information of at least one first medicine to be taken at each medicine taking time and a medicine storage grid in the corresponding intelligent medicine box, and for each medicine taking time, in response to the arrival of the medicine taking time, the medicine taking reminding text of each medicine taking time is obtained. According to the method, the medicine taking reminding voice of the medicine taking time is played, the indicator lamps on the medicine storage grids corresponding to the first medicines to be taken at the medicine taking time are controlled to be turned on, the target object is reminded of the information of the medicines to be taken through multiple reminding modes, the situation that a medicine taker forgets to take medicine is avoided, and the richness, fineness and accuracy of the reminding information are improved.

The invention discloses an information display method, device and equipment and a readable storage medium, and belongs to the technical field of electronic equipment. The information display method comprises the steps of determining first information according to user data; the user data comprises first data, and the first data is data of using the electronic equipment by a user; and displaying the first information.

A method and system for identifying wireless communication devices based on radio frequency fingerprinting is disclosed. The receiver extracts radio frequency fingerprint features from the wireless signals transmitted by the wireless communication device to identify its identity. The method includes: Step 1. Pre-training a feature extraction network and a classification head using known device label data; Step 2. Jointly estimating the number of known and unknown device categories in the mixed signal through feature space stripping and effective rank estimation; Step 3. An adaptation phase, freezing the anchor points of the known class classification head and constraining the anchor points of the unknown class to actively guide isolation; Step 4. Obtaining pseudo-labels by performing semantic geometric registration on unlabeled samples based on balanced entropy regularized optimal transmission; Step 5. Adaptively training the feature extraction network based on progressive manifold anchoring and reliable pseudo-replay; Step 6. Outputting the identity of the device to be identified using the trained feature extraction network and a unified classifier. This invention solves the stability-plasticity dilemma of known class forgetting and unknown class confusion, achieving accurate device identification and number estimation in dynamic environments.

The invention discloses an integrated gas source box with an independent turn-off function and an operation method thereof, and the integrated gas source box comprises a box body which is used for protecting internal installation parts and positioning the position. The integrated control assembly is arranged in the air source box, compressed air in the air source box can be reasonably controlled by the integrated control assembly, independent turn-off can be carried out, concentrated exhaust can be carried out, waste gas is prevented from remaining in the air source box to damage components, the convenience in use is improved, and the service life of the air source box is prolonged. Meanwhile, a portable connecting assembly is arranged on the integrated control assembly, operation during later maintenance of the integrated control assembly is facilitated, the maintenance efficiency is improved, and a tool storage assembly needed during later maintenance is arranged in the air source box in advance, so that necessary tools can be stored; and convenience is brought to workers during maintenance and overhaul, and tools needing to be used are prevented from being forgotten or lost in the working process.

The utility model relates to the field of grounding wire, concretely to a contact net pre -buried channel flexible grounding bow -shaped grounding wire structure, including steel wire rope, bow -shaped end and L -shaped end, the bow -shaped end and L -shaped end are detachable respectively connected in the both ends of steel wire rope, the bow -shaped end and L -shaped end are opposite respectively with steel wire rope rotation, the steel wire rope is flexible steel wire rope, the welding surface of bow -shaped end and L -shaped end is connected with pre -buried channel and contact net welding respectively, the bow -shaped end includes bow -shaped welding section and bow -shaped bolt section, the bow -shaped welding section is connected with pre -buried channel welding, the bow -shaped bolt section is detachable with steel wire rope connects, the utility model can make contact net and pre -buried channel grounding installation more convenient, avoid simultaneously two lining trolley to rise top after grounding welding welding slag burn or damage two lining trolley on already brushed good release agent.

The invention provides an inflatable paddle board which solves the technical problems that fins of an existing paddle board adopt a detachable connection mode, the fins and the paddle board are divided into two parts after being stored, and the situation that the fins are omitted and forgotten to be installed can occur during use and storage, and the inflatable paddle board can be widely applied to the field of surfing paddle boards. The device specifically comprises an inflatable paddle board, an inflation inlet is formed in the inflatable paddle board, a fin is arranged at the tail of the bottom face of the inflatable paddle board, the joint of the fin and the inflatable paddle board is integrally formed, and the fin is of a deformable flexible structure; boundary ribs are arranged on the frame of the fin, reinforcing ribs are arranged in the boundary ribs, the boundary ribs are made of deformable hard materials, and the reinforcing ribs are made of non-deformable hard materials; the boundary rib is composed of a first fixed rib, a second fixed rib, a first movable rib and a second movable rib which can define a fin shape, and the first fixed rib and the second fixed rib are each provided with a folding and unfolding sliding groove.

The embodiment of the invention provides a message pushing method and system based on burying point data and related equipment, belongs to the technical field of computers, and is suitable for the field of financial science and technology. The method comprises the following steps: firstly, in response to an operation prompt request, obtaining a current buried point website, operator information and target buried point data, obtaining current configuration information of the current buried point website, and further screening out reference configuration information from preset historical configuration information based on the current configuration information; and finally, generating an operation prompt message according to the reference configuration information and the target burying point data, and determining a client according to the operator information, thereby prompting the operation prompt message through the client. Therefore, according to the embodiment of the invention, the operation prompt message can be efficiently and accurately generated, and the operation prompt message can be accurately and timely transmitted to the operation personnel, so that the operation personnel can effectively receive and understand the operation prompt message.

The invention relates to the technical field of access control management, in particular to an access control system and method based on a face recognition technology, and the system comprises an access control linkage unit, and also comprises an image acquisition unit which is used for carrying out face image acquisition on a to-be-recognized user based on a camera according to a preset sampling period, and the face recognition unit is used for receiving the to-be-recognized user to obtain a face image sequence of the to-be-recognized user, and transmitting the face image sequence of the to-be-recognized user to the image preprocessing unit. Through the linkage of the face image acquisition unit, the image preprocessing unit and the face recognition unit, the safety and convenience of the access control system are effectively improved, the image acquisition unit automatically acquires the face image of the user to be recognized based on the preset period, and the face recognition unit quickly recognizes the image, so that the user experience is improved. The process not only improves the recognition speed, but also avoids the manual operation of traditional card swiping, password input and other modes, and simplifies the process of user access.

The embodiment of the invention provides a password setting method, an unlocking method, electronic equipment and a storage medium. The method comprises the steps of obtaining input first password information; generating one or more associated password images according to the first password information; receiving a selection operation for one or more associated password images; and extracting and storing the features of the selected target password image. According to the scheme of the embodiment of the invention, the method is faster, more convenient, more efficient, more diversified and more scene-based on guaranteeing the safety, is easy to memorize, and avoids the forgetting of passwords.

The utility model relates to the field of electronic cigarette discloses an air switch air channel structure of preventing condensate, including zinc alloy shell, PCB assembly, air adjusting device, POD oil storehouse base and data line storage mechanism, data line storage mechanism includes fixed box, the fixed box in top end fixed setting has micro motor, the micro motor output fixedly covers and sets up one no. In the utility model, the data line does not need to be carried additionally, the problem that traditional data line is lost or forgotten can be effectively avoided, after charging, the micro motor can be started to wind the data line neatly back to the outer end of the first binding roller.

The invention discloses a patient state information supervision method and related equipment. The method comprises the following steps: acquiring a voice record input request of a patient to which the equipment belongs through wearable intelligent terminal equipment; collecting effective voice information of the patient based on the voice record input request, wherein the effective voice information comprises diet types and diet amounts; querying the water content associated with the diet type, and evaluating the water intake of the patient based on the water content and the food intake. The problems that at present, a clinical nurse manually records the liquid inlet and outlet amount of a patient at a fixed time point, the consumed labor time is long, the water content of food is inaccurately estimated, when the condition of the patient changes, a physician cannot obtain completely accurate inlet and outlet amount data at the first time, the condition judgment time is delayed, and the accuracy is high can be solved. And even inappropriate medical advices can be given.

The invention belongs to the technical field of artificial intelligence, and relates to a target detection method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a playback image set and a pre-trained target detection model which comprises a front layer, a playback layer and a full connection layer; during forward propagation, inputting a training playback image to a front layer, extracting image features, obtaining an activation value corresponding to the training playback image from playback storage, connecting the image features and the activation value in a playback layer, inputting a connection data structure to a full connection layer, and determining a category result of a target object; and during back propagation, calculating a gradient according to a category result, updating a weight parameter of the full connection layer, and obtaining a trained target detection model. In addition, the invention also relates to a block chain technology, and data such as a target detection model can be stored in a block chain. According to the method and the device, disastrous forgetting of the model can be overcome, and accurate and efficient object detection can be realized on the edge equipment with limited resources.

The application discloses an intelligent network operation and maintenance management method based on user behavior analysis and belongs to the technical field of network operation and maintenance management. The method obtains original behavior log data of a user terminal, performs double alignment of a behavior baseline and an abnormal deviation degree to obtain a user behavior time sequence feature sequence; a user behavior analysis model including an input gating layer, a behavior memory evolution layer and an output mapping layer is constructed, the input gating layer splits the feature sequence into a steady-state behavior component and a transient behavior component; the behavior memory evolution layer differentiates and stores and fuses the two components through a competition mechanism to generate a behavior evolution state field; the output mapping layer performs nonlinear decoupling on the behavior evolution state field to output a user behavior abnormal risk measure and a behavior prediction path; and a potential risk node and an expected behavior trajectory are determined according to the user behavior abnormal risk measure and the behavior prediction path, and a dynamic access control strategy is generated. The method can improve the abnormal behavior recognition accuracy and the timeliness of dynamic adjustment of access control.

The utility model discloses a chronic disease recording device which comprises a reminder, a display screen, a timing module and a control module are arranged on the reminder, an alarm lamp and a buzzer are installed on the top of the reminder, a base plate is arranged below the reminder, a clamp is fixed to the top of the base plate, a recording chart is clamped in the clamp, and a marking piece is arranged on the recording chart. A storage box used for containing the marking piece is arranged on the back face of the base plate, and a control assembly used for controlling opening or sealing of the storage box is arranged on the back face of the base plate. When the condition of a patient in a certain day has abnormal fluctuation on data, the abnormal data are briefly recorded on the writing area, and the marking piece and the recorded page are pasted, so that a doctor can directly observe which day data is abnormal through the marking piece in the later period, and can directly find the recording page of the day, and therefore, the doctor can conveniently and quickly observe the abnormal condition of the patient. Detailed data comparison is facilitated, marking pieces do not need to be pasted on stable data, reading is not needed in the later period, and the working efficiency of doctors is improved.

The utility model discloses a hydraulic excavator safety lock device belongs to hydraulic excavator accessories field, including bottom plate, the top of bottom plate is installed with seat, the top fixedly connected with fixed frame of seat, the inner wall of bottom plate is provided with safety lock, the inside of fixed frame is provided with safety belt jackfield, convenient component, convenient component includes lock block and trigger block, lock block and trigger block all slidingly connect in the inside of fixed frame, the outside fixedly connected with horizontal pole of lock block, the top fixedly connected with the unlocking rod of trigger block, in the process of using, through the cooperation of above each device, in the process of using, through convenient component, can make the driving personnel after buckling up safety belt in the driving process, synchronous to safety belt locking also can realize the unlocking of hydraulic system, when the driving personnel leaves, need to unlock safety belt synchronous can lock safety lock, avoid personnel forget, improve convenience and security.

The utility model discloses a treadmill which comprises a treadmill body, an arc-shaped handrail is arranged on the treadmill body, the two ends of the handrail are respectively provided with an installation hole, and pluggable handheld monitors are arranged in the installation holes. The handheld monitor comprises a cylindrical shell, a power supply module, a monitoring module, a vibration module and a wireless signal transmission module, the power supply module, the monitoring module, the vibration module and the wireless signal transmission module are arranged in the shell, the monitoring module comprises a handheld heartbeat piece and a nine-axis attitude sensor and is used for monitoring the heart rate and the running posture, and the vibration module comprises a vibration motor and is used for vibration reminding. According to the utility model, the pluggable handheld monitor and the handheld monitor during running are adopted, so that the heart rate is monitored in real time under the condition that the rhythm of arm swinging during running is not influenced, the running posture can be monitored, the running posture is displayed in cooperation with the display screen, the posture of a body builder is corrected, and the exercise is healthier and more reasonable.

This invention belongs to the field of artificial intelligence technology, specifically relating to a cross-platform data interaction and sharing method and system based on artificial intelligence. It includes: constructing a health-aware edge cloud collaborative architecture to establish a dynamic health baseline for nodes; constructing a modified reward function with node health as a stability constraint, and using a multi-agent reinforcement learning model to train and obtain the optimal Nash equilibrium policy; decoding the optimal policy into an executable cross-platform data interaction planning graph through a policy interpreter network, and using game theory to coordinate resource conflicts; based on the planning graph, performing context-aware agent re-encryption and hierarchical homomorphic encryption, and establishing a power token chain to achieve fine-grained permission tracking; and performing closed-loop optimization of the policy and system self-evolution through multi-dimensional quality scoring, multi-layer source analysis, and elastic weight solidification technology; achieving synergistic gains in efficiency, security, and long-term stability of cross-platform data interaction.

The invention relates to a continuous Hash learning retrieval method and device for stream data, and the method comprises the steps: constructing a Hash retrieval model, introducing a continuous learning framework into the training process of the Hash retrieval model, firstly extracting a first feature from unlabeled training stream data based on self-supervised learning through a residual network, and then extracting a second feature from the unlabeled training stream data based on self-supervised learning; the convolutional neural network can quickly learn labeled sample data by using the first feature, and feeds back the learned second feature to the residual network to consolidate the residual network; the training stream data comprises first real-time stream data collected in real time and pre-stored stream data in a database, so that the model can avoid forgetting old pre-stored stream data knowledge while continuously learning new real-time stream data, and the continuous learning framework is also effectively connected with a hash function; the residual network and the convolutional neural network are used for continuous learning, and then extracted features guide learning of a hash function, so that the quality of generated hash codes is continuously improved, and the retrieval performance is further improved.

This application belongs to the field of artificial intelligence and healthcare technology, specifically disclosing a multi-memory large model enhancement method and system based on adaptive resonance theory. The method includes: obtaining memory units from the plot memory module based on a new medical dialogue event; processing the new medical dialogue event through a large model based on multi-source memory; multi-source memory is constructed based on information stored in memory units, working memory modules, and semantic memory modules; the plot memory module is used to: determine candidate plot prototypes by analyzing attribute similarity; determine whether there is resonance between the new medical dialogue event and the candidate plot prototype; if there is resonance, use historical dialogue events in the event set of the resonant plot prototype as memory units and update the resonant plot prototype; if there is no resonance, create a new plot prototype. This application enables effective memory management and improves the stability and accuracy of large model reasoning in medical dialogue scenarios.

The invention discloses a multi-modal intelligent osteoarthritis detection method based on loop optimization, a storage medium and electronic equipment. The multi-modal intelligent osteoarthritis detection method comprises the following steps of multi-modal data acquisition and preprocessing, modal-by-modal alternate training, test set division and uncertainty-based multi-modal fusion diagnosis; and finally, evaluating and optimizing the detection model. According to the method, a weighted aggregation mechanism based on uncertainty is introduced. In clinical scenarios, some modalities may have noise or missing. The model can automatically calculate the entropy value (uncertainty) of each modal prediction result, allocate a higher weight for a high-quality modal and allocate a low weight for a low-quality modal, so that misdiagnosis caused by poor quality of a single data source is avoided, and the diagnosis stability in a complex clinical environment is remarkably improved.