46results about How to "Improve inference accuracy" patented technology

Methods for detecting a common user of a plurality of user devices in a network and for detecting a common cohort for a plurality of users are disclosed. The methods comprise receiving a plurality of event records. Each of the event records corresponds to an event in a network and comprises a device identifier and event information. A correlation is then calculated between a first subset of the event records having a first device identifier and a second subset of the event records having a second device identifier. Based on the correlation, it is then calculated whether the first and second device identifiers relate to user devices associated with the same user, and whether the first and second device identifiers relate to user devices associated with users belonging to a common cohort.

The invention is aimed at providing a controlling device of a boiler equipment, which can eliminate the dangerous of dropping in locally optimal solution as much as possible even if the error between the model of a simulation controlled object and the action of the real controlled object exists, so that the controlled object can be operated and controlled steadily and safely. The invention calculates the tendency of the operation value based on the operation data, and calculates the error between the processing value calculated based on an equipment model and the detected value of the boiler equipment, and models the relation between the processing value and the operation order value, and then calculates the operation order candidate value by using the error value.

There is disclosed systems and methods to enhance reliability of measured position data. Measuring devices, such mobile phones equipped with location measurement elements (such as GPS, LBS, network location reporting, or tower location triangulation reporting) may collect various samples of positions where the device is believed to be located at particular moments in time; however such measurements often vary even if the device is not moving because of device inaccuracy, atmospheric conditions, obstructing buildings, and the like, making it difficult to determine whether such devices are actually stationary or are in motion over predetermined time periods. Systems and methods of the present invention provide for enhanced accuracy of position data by selectively merging varying location positions that are attributable to noise or accuracy deviations, and providing an enhanced assessment of actual device position.

This invention provides a gas concentration concluding method of a boiler equipment, making use of a neural network to conclude the concentration of the gas component exhausted from the boiler equipment, and making use of the neural network used by a plurality of neural networks prepared for each coal category and the neural network for judging the coal category proportion to conclude. The invention further provides a gas concentration concluding apparatus of the boiler equipment, making use of the neural network to conclude the concentration of the gas component exhausted from the boiler equipment, a, comprising a summate which makes use of the neural network used by a plurality of neural networks prepared for each coal category, the neural network for judging the coal category proportion and the mixed proportion concluded by the neural network for judging the coal category proportion to do weighted average to the concluded value of the gas concentration exported a plurality of neural networks prepared for each coal category.

The invention provides a method and a system for diagnosing equipment faults through a knowledge graph and a Bayesian network. The system for diagnosing the equipment fault through the knowledge graphand the Bayesian network comprises an input module and a diagnosis module, wherein the input module is used for inputting various data for recording maintenance methods and maintenance history; an OCR module used for identifying characters in the various patterns; an entity identification module used for processing the structured content in the document and identifying the types of the charactersin the structured content; and a professional dictionary module which comprises professional vocabularies and a stop word dictionary, wherein the stop word dictionary comprises some common words without entity meanings. According to the method and the system for diagnosing the equipment fault through the knowledge graph and the Bayesian network, the knowledge graph and the Bayesian network are established, so that the algorithm can dynamically learn, the reasoning precision of the algorithm is continuously improved, various probability data are updated, and the precision of future fault diagnosis is greatly improved.

The invention provides a welding channel emulate device of electric arc welding, which utilizes heat conduction computation to speculate the welding channel shape of electric arc welding and can perform the simulation corresponding to the type of welding power source used in construction. The device includes: a heat source model HSM which computes the input heat from electric arc to a melting pool, a melting pool model MPM which computes the shape of the melting pool via the heat conduction computation according to the input heat qc in order to speculate the shape of welding channel, and a welding power source simulation module PSS which modularizes the welding power source outputting transient welding current; in addition, the heat source model HSM comprises: an electric arc load simulation part which modularizes the electric arc load and takes the transient welding current io as input, takes transient welding voltage vo as output; the transient welding voltage vo is input to the welding power source simulation module PSS as a feedback signal, meanwhile, the input heat qc is computed in accordance with the transient welding current io and the transient welding voltage vo.

The invention provides a host vehicle position estimation device that can improve the accuracy in estimating the host vehicle position on the map. The host vehicle position estimation device 100 configured to detect objects of each of which a position on a map is set in advance using a vehicle-mounted sensor 2, and configured to use the objects for estimating a host vehicle position which is a position of the host vehicle on the map, the device including: a object database 5A configured to store information on the detected objects in which the position of each of the objects on the map and anobject-detectable area are associated with each other; a measurement position acquisition unit 11 configured to acquire a measurement position which is the position of the host vehicle on the map, measured by a vehicle-mounted GPS receiving unit 1; a candidate object recognition unit 13A configured to recognize a candidate object which is associated with the object-detectable area including the measurement position among the objects, based on the object-detectable area in the information on the detected objects and the measurement position; and a host vehicle position estimation unit 14 configured to estimate the host vehicle position based on a position of the candidate object on the map and a result of detection performed by the vehicle-mounted sensor.

The invention discloses a customized deep neural network model compression method and system based on cloud-edge cooperation, and the method comprises the steps of carrying out the classification of concerned classification and unconcerned classification of a network model based on a high-precision network model, trained for a long time, of a cloud end and the personalized demands of a user; and then, in combination with the concerned classification, the data set of the user and the model compression ratio, through a neural network model compression method based on knowledge arrangement, lightening the model, so that the scene requirement that the user edge node resources are limited is met, the model reasoning speed is increased, and meanwhile, certain model accuracy is ensured. Accordingto the model training method based on user attention classification requirements and the knowledge distillation technology, verification is carried out based on a public data set in a picture classification scene.

An Attention LSTM-based reinforcement learning Agent knowledge reasoning method can effectively memorize a knowledge graph reasoning path through a bidirectional long-short-term memory network, and meanwhile, an attention mechanism weights the state of the memory path to obtain a state needing to be concerned, inhibits an invalid state, realizes memory path screening, and the problem that knowledge reasoning of a reinforcement learning Agent cannot effectively memorize a reasoning path is effectively solved. The LSTMAttention network model is used in the reinforcement learning Agent for relation path feature extraction, meanwhile, a return mechanism is optimized, and the reasoning precision of a knowledge graph reasoning algorithm based on reinforcement learning under multiple reference data sets is effectively improved.

A host vehicle position estimation device includes a target recognition unit configured to recognize a relative position of the vertical position target relative to a host vehicle on the basis of a detection result of an external sensor of the host vehicle, an amount-of-position error calculation unit configured to calculate an amount of position error, an amount-of-steady error calculation unit configured to calculate an amount of steady error from a distribution of the amount of position error when it is determined that the distribution of the amount of position error satisfies a steady condition, and a host vehicle position estimation unit configured to perform estimation of a vertical position of the host vehicle using the amount of steady error when the relative position of the vertical position target relative to the host vehicle is not recognized.

The invention relates to the technical field of games, in particular to an information processing method and device in a game and a computer medium and electronic equipment for achieving the method. The method comprises the steps of for first touch operation, determining at least one piece of trace information; for second touch operation, associating the trace information with a preset informationcollection container; for third touch operation which acts on the preset information collection container, providing a preset collection display area, and displaying the collected trace information in the preset collection display area, wherein the collected trace information is the trace information associated with the preset information collection container; for fourth touch operation which acts on the preset collection display area, generating a relation instruction used for instructing a relation of the collected trace information. According to the technical scheme, by means of the information processing method in the game, the efficiency of collection and inference by means of traces can be improved, and meanwhile, an inference process can be embodied for a user visually, so that theinformation processing method in the game benefits improvement of the inference accuracy.

The present invention provides an air conditioner, which uses a plurality of human detection sensors to divide the area requiring air conditioning into a plurality of areas according to the positions of people in the area requiring air conditioning, and sets the frequency of each area according to the frequency of people. regional characteristics. Based on the response result of the human body detection sensor within the first predetermined time period, a judgment condition indicating human presence is set in each area characteristic, and whether there is a person in each area is estimated based on the set judgment condition. According to the response result of the human body detection sensor in the second specified time, when a person is detected in the first area, it is inferred that there is a person at a time interval shorter than that of the second area; Since no one is estimated at a long time interval, the time required for inferring presence in the first area is set shorter and the time required for inferring no one is set longer in the first area than in the second area.

The invention relates to a wear amount estimation device, a wear amount estimation method, and a wear amount estimation program for improving the estimation accuracy of the wear amount of a brake pad of a vehicle. The amount of wear function represents the amount of wear as a function of vehicle speed, brake pressure, and brake duration. When the vehicle is braked, the wear amount estimation device executes a wear amount calculation process for calculating a wear amount corresponding to the vehicle speed, the brake pressure, and the brake duration by using the wear amount function. In addition, the wear amount estimation device estimates the temperature of the contact portion of the brake pad in contact with the brake rotor on the basis of the vehicle speed, the brake pressure, and the brake duration. On the basis of the temperature of the contact part, the wear amount estimation device updates temperature history information indicating at least the temperature history of the contact part. In the wear amount calculation process, the wear amount function is variably set on the basis of the temperature history of the contact portion indicated by the temperature history information obtained at the time of the last braking.

Provided are a learning method and device of a model used for processing of automatically generating a text from an image, a program, a learned model, and a text generation device. In a learning method according to an embodiment of the present disclosure, a learning model including a feature amount extraction unit that extracts an image feature amount from an image, a text generation unit that generates an estimated text from the image feature amount, and a statistic estimation unit that generates an estimated image statistic from the text is used, and a combination of a first image, first metadata, and a first text is used as learning data. A first error of a first estimated text estimated from a first image feature amount of the first image, a second error of a first estimated image statistic estimated from the first text, and a third error of a second estimated image statistic estimated from the first estimated text are calculated, and a parameter of the learning model is updated.

The invention discloses an SNN training method and device, a storage medium, a chip and an electronic device. In order to solve the technical problems of instability, low efficiency and poor precision when the spiking neural network is trained, a time constant or an attenuation coefficient in the spiking neural network is trained through a space domain and time domain error return allocation strategy. The technical scheme disclosed by the invention not only considers the spatial domain dependency of the pulse neural network, but also considers the time domain dependency between pulses, so that the technical problem of how to efficiently and stably train the time constant or the attenuation coefficient in the space-time domain is solved; the technical effects that the SNN can be trained more quickly and stably, and the brain-like chip deployed with the SNN configuration parameters obtained through training has higher reasoning precision performance are achieved. The method is suitable for the fields of pulse neural network training and brain-like chips.

The invention discloses a rough set-confidence rule extraction method and system, and a medium. The implementation steps of the method comprise: establishing a knowledge expression system for an object of a confidence rule base and a data set thereof, determining the number of initial reference values, setting the initial width of a reference value interval as a fixed value, carrying out equivalent class division, calculating confidence, and extracting a confidence rule; Reasoning a confidence rule base established by the test set based on the extracted confidence rule; and optimizing the number of reference values and the reference value interval according to the reasoning result to obtain the optimal number of reference values and the reference value interval, then performing equivalentclass division based on the optimal number of reference values and the reference value interval, and calculating confidence to extract a confidence rule. According to the reasoning result, the initially determined attribute index is optimized, so that the optimal attribute index value can be obtained, and the reasoning precision of the confidence rule base system can be improved.

The invention discloses a deep learning model reasoning method, system and device, and a computer medium. The method comprises the following steps: determining channel quantization parameters of each channel in a target deep learning model; determining a branch quantization parameter of the model branch based on the channel quantization parameter according to a relationship between a channel in the target deep learning model and the model branch; quantizing the target deep learning model based on the branch quantization parameters to obtain a quantized deep learning model; and reasoning based on the quantized deep learning model to obtain a reasoning result. In the application, the channel quantization parameter of each channel in the target deep learning module is determined, fine-grained quantization parameter determination is realized, the branch quantization parameter of the model branch is determined based on the channel quantization parameter, reasoning is performed based on the quantized deep learning model, and the reasoning result is obtained. The deep learning model reasoning based on the fine-grained channel quantization parameters is realized, and the reasoning precision of the deep learning model is improved.

There is disclosed systems and methods to enhance reliability of measured position data. Measuring devices, such mobile phones equipped with location measurement elements (such as GPS, LBS, network location reporting, or tower location triangulation reporting) may collect various samples of positions where the device is believed to be located at particular moments in time; however such measurements often vary even if the device is not moving because of device inaccuracy, atmospheric conditions, obstructing buildings, and the like, making it difficult to determine whether such devices are actually stationary or are in motion over predetermined time periods. Systems and methods of the present invention provide for enhanced accuracy of position data by selectively merging varying location positions that are attributable to noise or accuracy deviations, and providing an enhanced assessment of actual device position.

The embodiment of the invention provides a method for calculating the angular deformation of a ship complex hull plate formation space. Through partitioning a machined curved surface and a target curve surface and calculating an angular deformation between corresponding blocks, at the same time, through carrying out precision calibration on each block, the partitioning of the machined curve and the target curve is more accurate, an error is in an acceptable error range, thus the calculation of the angular deformation is more accurate, the process parameter reasoning accuracy can be obviously improved, the hull plate formation efficiency is improved, the method has a good practical application value, and a technical problem of the low calculation precision of a method of measuring front and back ship outer sides is solved. In addition, the embodiment of the invention provides a device for calculating the angular deformation of a ship complex hull plate formation space.

An object of the present application is to provide a stage of charge (SOC) estimation device and a charge state estimation method that can improve the estimation accuracy of the SOC even if the discharge or charge of the secondary battery continues. In the SOC estimation apparatus, a measurement value acquiring unit obtains a voltage measurement value, a current measurement value, and a temperature measurement value of the second battery; an estimation value calculating unit calculates a SOC estimation value of the secondary battery based on the current measurement value; a voltage monitoringunit determines whether or not the voltage measurement value of the secondary battery has reached a predetermined voltage threshold during charge or discharge of the secondary battery; and in a case where the voltage measurement value of the secondary battery is determined to have reached the voltage threshold, a compensation unit corrects the SOC estimation value based on the current measurementvalue and the temperature measurement value that are obtained at a correction time point at which the voltage measurement value of the secondary battery has reached the voltage threshold.

The invention provides an intersection signal timing parameter inferring method based on checkpoint detection data, and belongs to the technical field of traffic signal timing research. Inferred intersection signal timing parameters comprise phase sequence, cycle time and phase green light time. According to the method, on the basis of the checkpoint detection data, the phase sequence is inferred based on vehicle passing time sequence information according to the geometrical characteristics of an intersection, and the phase green light time is inferred by introducing phase weight and phase average time information; and the mutual complementation of information among each entrance lane, a non-single direction and each release direction of the intersection is considered, actual demands are fully considered, a logic chain is complete, the inference accuracy is high, the model has high robustness, and the scene transportability is high. The method can be suitable for actual checkpoint vehicle passing data and other data with identity tag detection; and in addition, the method can further carry out problem discovery and calibration on clocks of a signal control system and a checkpoint system.

This invention provides a gas concentration concluding method of a boiler equipment, making use of a neural network to conclude the concentration of the gas component exhausted from the boiler equipment, and making use of the neural network used by a plurality of neural networks prepared for each coal category and the neural network for judging the coal category proportion to conclude. The invention further provides a gas concentration concluding apparatus of the boiler equipment, making use of the neural network to conclude the concentration of the gas component exhausted from the boiler equipment, a, comprising a summate which makes use of the neural network used by a plurality of neural networks prepared for each coal category, the neural network for judging the coal category proportion and the mixed proportion concluded by the neural network for judging the coal category proportion to do weighted average to the concluded value of the gas concentration exported a plurality of neural networks prepared for each coal category.

The invention provides a prostate nodule ultrasonic-assisted diagnosis system, a storage medium and electronic equipment, and relates to the technical field of ultrasonic-assisted diagnosis. The reasoning precision of a single extended belief rule base system is easily influenced by the quality of a training data set, and the Bagging algorithm is utilized to construct a plurality of basic extended belief rule base systems, so that the systems can complement each other, and the overall reasoning precision is improved. The attribute weight and the rule weight are objectively determined by constructing the optimization model, the subjectivity of the result can be effectively avoided, and the result obtained by the system is more objective and accurate. With the system adopted, the prostate ultrasonic diagnosis precision can be effectively improved so as to assist doctors in diagnosing prostate pathology results.

An information processing apparatus that performs machine learning of a learning model for improving the accuracy of inference of an image even in a case where the characteristics of an image are changed by correction applied to the image. The image processing apparatus includes an image processor configured to perform image processing on a training image and a teacher image according to each of a plurality of correction conditions. Machine learning of the learning model is performed using the training image and the teacher image. A plurality of parameters obtained by performing the machine learning of the learning model are stored in association with the plurality of correction conditions, respectively.

Provided are a machine learning system and method, an integration server, an information processing apparatus, a program, and an inference model creation method capable of solving a problem caused by non-uniform quality of learning data in federated learning and improving an inference accuracy of a model. Each of a plurality of client terminals classifies data stored in a medical institution based on a data acquisition condition and classifies learning data into each data group acquired under the same or a similar acquisition condition. Each client terminal executes machine learning of a learning model for each learning data group classified into each condition category and transmits each learning result and condition information to an integration server. The integration server integrates received learning results for each condition category to create a plurality of master model candidates and evaluates the inference accuracy of each master model candidate.

The invention discloses a step-by-step inference method for business relationships of Internet autonomous domain systems, which comprises the following steps: firstly, based on a valley-free path principle, according to top-level autonomous domain system information and autonomous domain system degree information, carrying out principle inference on relationships between adjacent autonomous domain systems of each path in a path table; secondly, performing re-judgment on an autonomous domain system relationship with conflicts in principle inference by using a mode of combining probability inference and principle inference, preferentially determining a result with a high global inference proportion, and updating the principle inference according to a new result; and for the autonomous domain system relationship of which the inference result is still not determined, inference is carried out according to the autonomous domain system degree information, and finally the commercial relationship between the adjacent autonomous domain systems on the path in the routing table is obtained. According to the method, the accuracy of deducing the types of the business relations between the autonomous domain systems is effectively improved, and the method is suitable for Internet routing security and business relation analysis of the autonomous domain systems.

A data padding method includes adding at least one padding column or at least one padding row to a data matrix. One of a plurality of elements of the at least one padding row or the at least one padding column is different from another of the plurality of elements.

The invention provides a network information diffusion source inference method based on a differential pre-solution set and belongs to information technology processing field. The method comprises steps of using G = (V, E) to represent a connection relationship between the network nodes, wherein V represents a network node set, e represents a set of network edges; selecting a difference pre-solution set S from the difference pre-solution sets; establishing diffusion source characteristics, collecting cascading information, judging the module | Ic | of an index set, if the module | Ic | is greater than a preset value, extracting a characteristic vector of cascading C, adjusting the diffusion source characteristics of each node v belonging to V, calculating norms of all nodes v belonging toV one by one, and finding out the node with the minimum norm as a source. According to the method, the data source can be actively optimized and selected, the quality of input data is improved, and the source inference precision is greatly improved.

The invention relates to a model quantification method and device, electronic equipment and a storage medium. The method comprises the steps that a quantification model set is acquired; the quantization model set comprises at least two quantization models; the quantization model is obtained by quantizing at least one target network layer in a to-be-quantized model; determining fitness corresponding to each quantitative model; the fitness is used for representing performance improvement of the quantitative model; the performance improvement of the quantitative model is determined according to the difference between the model performance of the quantitative model and the model performance of a preset model; the preset model is obtained by quantizing each network layer in the to-be-quantized model; the model performance comprises model speed and model precision; and performing genetic algorithm optimization on the quantization strategy parameters corresponding to the quantization models according to the fitness corresponding to the quantization models to obtain a target quantization model. According to the invention, the quantification effect of the neural network model can be improved.