60 results about "Model sharing" patented technology

A computer program product includes machine readable instructions for managing data items, the instructions stored on machine readable media, the product including instructions for: initializing a plurality of base models; minimizing a joint loss function to select models from the plurality for a plurality of labels associated with the data items; and at least one of sharing and combining the selected base models to formulate a composite classifier for each label. A computer system and additional computer program product are provided.

The invention discloses a random convolutional neural network-based high-resolution image scene classification method. The method comprises the steps of performing data mean removal, and obtaining a to-be-classified image set and a training image set; randomly initializing a parameter library of model sharing; calculating negative gradient directions of the to-be-classified image set and the training image set; training a basic convolutional neural network model, and training a weight of the basic convolutional neural network model; predicting an updating function, and obtaining an addition model; and when an iteration reaches a maximum training frequency, identifying the to-be-classified image set by utilizing the addition model. According to the method, features are hierarchically learned by using a deep convolutional network, and model aggregation learning is carried out by utilizing a gradient upgrading method, so that the problem that a single model easily falls into a local optimal solution is solved and the network generalization capability is improved; and in a model training process, a random parameter sharing mechanism is added, so that the model training efficiency is improved, the features can be hierarchically learned with reasonable time cost, and the learned features have better robustness in scene identification.

The invention discloses a natural language processing model training method and device. The method is applied to a plurality of natural language processing models with different task targets. A plurality of natural language processing models share a part of hidden layers, and the method comprises the following steps: fusing label information of training samples corresponding to all tasks and representing the fused label information as a multi-label matrix, calculating a hit probability according to a task target vector output by the model and the multi-label matrix, and calculating a single task loss value according to the hit probability; calculating a multi-task loss value according to each single-task loss value; and finally, adjusting parameters of each model according to the multi-task loss value. In the method, a plurality of natural language processing models are combined for learning; shallow feature representations of a plurality of tasks are shared by sharing part of hidden layers, gradients are propagated back at the same time to help the hidden layers to escape from a local minimum value, fusion of a plurality of labels and vector representations of the labels help increase the expression generalization ability and the representation ability of the labels, and then the accuracy and the generalization of each model can be improved.

Disclosed is a method and a system a method and a system for auto learning, artificial intelligence (AI) applications development, and execution. Various applications or operations may be associated with training environment-agnostic AI models, automated AI app application performance monitoring, fault, quality and performance remediation through prediction of failures or suboptimal performance, privacy and secure AI training and inference mechanism for data and AI model sharing between untrusted parties, and building auto learning applications that can automatically learn and improve.

The embodiment of the application discloses a model transfer method based on spectral data. The method includes measuring indexes of samples and collecting spectra; pretreating the spectra; dividing aplurality of the calibration samples into correction samples and prediction samples; preforming model transfer from a main instrument to a target instrument, namely, forward transfer; and alternatively, performing model transfer from the target instrument to the main instrument, namely, backward transfer. Compared with the traditional method, the method has the following advantages that: the interference of noise and irrelevant information can be suppressed through the correlation between the collected spectral data of the samples and the spectral data structure; two-way model transfer of themain instrument and the target instrument can be realized, and whether the instrument types of the main instrument and the target instrument are consistent or not is unrestricted, and whether collected spectral data points are consistent or not is unrestricted, so that the method is of great significance for model sharing of different types of spectrometers with the same spectral method and improving the applicability of the model.

Integrated modeling and symbolic manipulation is leveraged for the derivation, construction, maintenance, and reuse of application-independent models. Related models created for different applications share a common ancestry and maintain model consistency while enabling the models to share information about the process. Environment independent proper ancestor models (PAMs) are created for generic components of a process, such as a distillation or flash column. Models underlying an application may be comprised of various representations of multiple such processes corresponding to many different physical components. PAMs contain symbolic representations of different sub processes that occur within a process or component. A user makes assumptions about the component, by indicating how sub processes are to be considered or not considered. PAMs are then modified in accordance with the assumptions to derive a specific environment model (SEM).

The present invention relates to an audio quality recovery system based on GAN. The system comprises a model sharing block module, a generation network model module, a distinguishing network module and a sequence recombination module; the model sharing block module is mainly used to perform feature extraction for time-domain signals without frequency domain processing due to avoidance of information loss and extract the features to a high-level unit; the generation network model module employs a high-level extraction unit to perform analysis and reconfiguration; the distinguishing network module is configured to continuously perform adversarial training with the generation network model to continuously improve a generation effect; and the sequence recombination module analyzes the sequenceweighing recombination of the final generation output by the network. The audio quality recovery system based on the GAN can generate more vivid audio signals.

The invention provides a method for coordinating and sharing a transformer substation end and a dispatching end model based on an E language. The method comprises the following steps: (1) producing an SCD (security coding device) file; (2) reading the SCD file and analyzing the SCD file; (3) according to the read SCD file information, building an E language model file; (4) performing model checking for the E language model file; transmitting the checked E language model file to a dispatching automatic system; introducing the checked E language model file to the dispatching automatic system by an E language model file introducing tool in the dispatching automatic system, otherwise, entering step 5; and (5) modifying the E language model file which is not qualified. With the adoption of the method for coordinating and sharing the transformer substation end and the dispatching end model based on the E language, grid model sharing between the transformer substation end and the dispatching end can be realized, and the repeating, dividing and isolating modeling problem between the transformer substation end and the dispatching end is avoided.

The invention discloses a model sharing method between a power distribution main station and a terminal. The method comprises the following steps of comparing an update time t1 with a last update time t2, and if t1>t2, updating a power distribution terminal model; obtaining a to-be-updated model; performing mapping association between an IEC61968 / IEC61970 model in the power distribution main station and a terminal model based on the IEC61850 standard; comparing with a current mode after the power distribution terminal receives an updated file; uploading a model comparison result by the power distribution terminal; sending an operation instruction by the power distribution main station according to the comparison result; conducting operation by the power distribution terminal according to the operation instruction; and uploading a final update time of the power distribution terminal model by the power distribution terminal. The invention utilizes the mapping function of the IEC61968 / IEC61970 and IEC61850 models and provides a powerful guarantee for rapid and accurate position and processing of power distribution network faults.

The invention discloses a cloud simulation service method comprising the steps of: virtualizing hardware resources in a cloud simulation architecture; virtualizing software resources in the cloud simulation architecture; establishing a unified management platform in the cloud simulation architecture to manage the virtualized hardware resources and the virtualized software resources; and constructing a the professional model and a solver in the cloud simulation architecture. The method has the following advantages of using model sharing service, resource flexible scheduling and solver parallel computing as a core, scheduling tasks and distributing computing resources on demand, providing the professional model and the solver for engineering staff in the form of services by using service middleware, configuring on demand, providing a flexible and reliable computing platform for enterprises; and achieving high data security and high data reliability.

The invention discloses a method and a device for realizing sharing of a contextual model, so as to solve the problem in the prior art that user terminals can not share the contextual model. The method comprises: a sharing server receives a contextual model sharing request transmitted by a first user terminal, wherein the contextual model sharing request carries contextual model information requested for sharing and information of other users sharing the contextual model; and the sharing server transmits a contextual model synchronization request to user terminals corresponding to the information of the other users according to the contextual model sharing request, wherein the contextual model synchronization request carries contextual model information, so that the sharing server requests the user terminals to set the contextual model for communication according to the contextual model information. By adoption of the technical proposal, the method and the device can realize contextual model sharing between the user terminals and make the contextual model more flexibly set by the users.

The invention discloses a power grid model sharing method applicable to distributed computation of a power system. The method comprises: constructing a power grid physical model applicable to multi-stage scheduling integrated distributed computation; performing local and global verification on a distributed integrated power grid model sequentially; performing distributed storage and redundancy backup on the power grid model that passes the verification; enabling the power grid physical model that passes the verification; and integrating a large power grid distributed network topology analysis function and a distributed state estimation function to a model management center, and accessing to the power grid physical model and a computation model according to different demands of applications. According to the method, existing storage resources of distributed deployment of each scheduling center are fully used, and physical distributed storage and logic centralized management are performed on the power grid physical model and the computation model by using a distributed data deployment policy, so that the power grid model is constructed. Based on a distributed database technology and a distributed file system technology, full-network instant sharing of the power grid physical model and the computation model is implemented.

The embodiment of the invention provides a training method of a text translation model and a text translation method and device, and the training method comprises the steps: obtaining a parallel corpus set, and enabling each parallel corpus in the parallel corpus set to comprise a source language text, a machine translation text, a post-edited translation text and a revised translation text; sequentially training a first translation model, a second translation model and a third translation model according to the parallel corpus set, taking the trained third translation model as the text translation model, enabling the first translation model, the second translation model and the third translation model to share part of model parameters, and enabling the model parameters of the previous translation model after the training not to change along with the training of the subsequent translation model. According to the training method provided by the embodiment of the invention, a better model generalization effect can be obtained, the generalization ability of the model is improved, and finally, the translation result obtained by training is more accurate.

A novel electronic system provides fast three-dimensional model generation, social content sharing of dynamic three-dimensional models, and monetization of the dynamic three-dimensional models created by casual consumers. In one embodiment, a casual consumer utilizes a dedicated real-time 3D model reconstruction studio with multiple camera angles, and then rapidly create dynamic 3D models with novel computational methods performed in scalable graphics processing units. In another embodiment, uncalibrated multiple sources of video recording of a targeted object are provided by a plurality of commonly-available consumer video recording devices (e.g. a smart phone, a camcorder, a digital camera, etc.) located at different angles, after which the uncalibrated multiple sources of video recording are transmitted to a novel cloud computing system for real-time temporal, spatial, and photometrical calibration and 3D model reconstruction. The dynamic 3D models can be uploaded, listed, and shared among content creators and viewers in an electronic sharing platform.

The invention provides a multi-agent grid fault diagnosis system and method based on a blackboard model. A task planning Agent of the system divides a multi-agent system; a first information acquisition Agent acquires relay protection information of a power grid in real time, a second information acquisition Agent acquires circuit breaker information of the power grid in real time, and the information is transmitted to a blackboard model sharing platform; the blackboard model sharing platform determines whether the transmission data of the relay protection information and the circuit breaker information are fault data by setting a switching function, updates the received fault data, and obtains the accurate fault data; an information transmission Agent compares and superimposes the received accurate fault data one by one, and obtains the final fault data; a topology Agent determines the topological structure of a fault region; and a fault diagnostic Agent conducts a fault diagnosis to diagnose faulty elements using a Petri network based on the fault data and the topological structure of the fault region.

Provided are a method of upgrading software by using a common module shared by a plurality of models, and a digital broadcast receiver therefor. The method includes receiving a digital broadcast signal which contains a common module shared by a plurality of models of the digital broadcast receiver, separately from a unique module of the digital broadcast receiver; determining whether to upgrade each module of the digital broadcast receiver, when it is determined that software upgrading is needed, according to predetermined update information; and collecting DownloadDataBlocks (DDBs) for a module, which is to be upgraded, from the received digital broadcast signal, in order to form a module. The common module is transmitted in the form of a group different than a group to which the common module belongs or in the form of another group, thereby saving bandwidth for software upgrading.

The invention discloses a model sharing and rendering method based on Alembic and an electronic terminal. The method comprises the following steps: selecting a locally stored model file and importingthe model file into an animation component; the animation component obtains the original model and the replication model of the original model according to the model file. In the animation component,copy model is selected as the deformation basis, and the original model and copy model are fused and deformed to obtain the target model. whether to cache the target model is judged; after judging thecache target model, the abc cache file of the target model and the rendering file matched with the abc cache file are exported through the animation component. The rendering file at least includes the material information and texture mapping information of the target model. The invention can avoid that the original model cannot be matched with the geometric result of the cache because of the subsequent modification of the original model, on the other hand, the original model can be quickly restored through the combination of the derived material information and texture mapping information andthe abc cache file, and the operation flow is simplified.

The invention discloses a design model sharing system in a cloud manufacturing environment. The design model sharing system comprises a design model information publishing module, a design model demand publishing module, a design model discovery module, a design model optimization module, a fund management module, a transaction management module and a service evaluation module. By applying the design model sharing system, the effects that repeated design is not needed in similar manufacturing enterprises and partial or whole of a design model can be directly obtained from the network can be realized, the design speed is accelerated, and the modular design is achieved; the design model sharing system is conductive to fully mining the value of internal data of the enterprises, converting enterprise product design models into data resources for sale, and realizing the reuse of the design models; and by applying the design model sharing system, a commercial network established by a plurality of design manufacturers can be formed, a transaction platform used for selling the design models is established for designers, and a service platform used for purchasing or selling the design models online is provided for various enterprises.

The disclosed embodiments provide a system for processing data. During operation, the system obtains a hierarchical representation containing a set of namespaces of a set of features shared by a set of statistical models. Next, the system uses the hierarchical representation to obtain, from one or more execution environments, a subset of the features for use in calculating the derived feature. The system then applies a formula from the hierarchical representation to the subset of the features to produce the derived feature. Finally, the system provides the derived feature for use by one or more of the statistical models.

The invention discloses an enterprise-level model management method and system, equipment and a storage medium. The method mainly comprises the steps: building an enterprise-level model asset library,and registering all models and related documents in the same management platform; monitoring the working state and the performance index of each model in the asset library in the life cycle in real time according to a pre-configured instrument board; carrying out model task processing, such as model requirement initiation, model development, model verification, model implementation and deployment, model approval, model online, continuous verification and monitoring after production and the like, on a universal model workflow according to different model task requirements. According to the invention, all the models are registered to the same model management platform; model sharing and multiplexing are carried out in the platform according to authorities of different roles, indexes or states of any one or more models are presented through an instrument board, and centralization, process and unification of model management can be realized through processing of different sub-tasks in ageneral model workflow association model life cycle.

The invention provides a multi-adaptive CIME power grid model sharing method. The method comprises the steps of S1, mapping CIME power grid model files into relation model files; S2, making table names in the relation model files and class names in CIM / RDFS model files correspond, and making the domains in the relation model files and the attributes in the CIM / RDFS model files correspond to form sharing model configuration files; S3, defining the sharing mode used when a CIME model performs model sharing in a current system to form model sharing scene files; S4, constructing a CIM / RDFS-based power grid model and performing correlation according to the correlation rules of various kinds of objects of the CIM-RDFS-based power grid model. The method maintains the high efficiency of CIM / E model files in a model interaction process and combines the characteristics of convenient use, flexible model extension and reusability in CIM / RDF Schema; the multi-adaptive model sharing method is provided for engineering application.

The invention provides a method for generating a transformer station anti-error locking system communication model. The transformer station anti-error locking system communication model is processed and generated by taking an intelligent electronic device of a measurement and control apparatus as a basis. The method comprises one or more from a first step, a second step and a third step: step one, generating a switch locking module and a knife locking model; step two, generating a net door locking model and a ground wire locking model; step three, generating a programmed operation ticket transmission control model; and step four, generating a data set and a report control block model. According to the invention, the modeling problem of an anti-error locking device, the modeling problem of a programmed operation ticket transmission model and the problem of data interaction among integrated automation systems which are not related to by a conventional IEC61850 standard system can be solved, model sharing between an anti-error system and an automation system is realized, interaction models of common operation and programmed operation of local monitoring, centralized control, scheduling systems and an anti-error locking system are unified.

The invention discloses a data sharing method based on a block chain and federated learning, and the method comprises the steps: building mutually trusted block chain nodes into a team, receiving a request task, and then selecting a team meeting a credit rating requirement to respond to the request task; after the data sharing task is received, the nodes in the team meeting the credit rating requirement are used for training a verification model until the verification model reaches preset accuracy or reaches the maximum training time, and model sharing is achieved to protect privacy of a data provider; a model training process is packaged locally, a consensus algorithm based on node contribution reaches a consensus among block chain nodes, and credit reward is performed on a team meeting a credit rating requirement, so that each training process in a data sharing process is recorded to ensure that a data provider provides high-quality data; the credit reward is carried out after the consensus is reached, the credit rating can be updated in time, the reliability of the credit rating is ensured, and the privacy protection problem of data in the Internet of Things is relieved.

The invention discloses an intelligent production state monitoring system comprising a data acquisition module, a data processing and display terminal, a server and a cloud. By installing an acquisition board on a three-phase electric-driven equipment motor to collect the equipment processing data in real time, and establishing models of the production process and abnormal process by a learning algorithm of the data processing and display terminal, real-time monitoring, recording feedback and abnormal warning of the status of the production system including production equipment, products and production personnel are realized, and services such as communication, model sharing, data storage and query are provided for each data processing and display terminal through the server.

A novel electronic system provides fast three-dimensional model generation, social content sharing of dynamic three-dimensional models, and monetization of the dynamic three-dimensional models created by casual consumers. In one embodiment, a casual consumer utilizes a dedicated real-time 3D model reconstruction studio with multiple camera angles, and then rapidly create dynamic 3D models with novel computational methods performed in scalable graphics processing units. In another embodiment, uncalibrated multiple sources of video recording of a targeted object are provided by a plurality of commonly-available consumer video recording devices (e.g. a smart phone, a camcorder, a digital camera, etc.) located at different angles, after which the uncalibrated multiple sources of video recording are transmitted to a novel cloud computing system for real-time temporal, spatial, and photometrical calibration and 3D model reconstruction. The dynamic 3D models can be uploaded, listed, and shared among content creators and viewers in an electronic sharing platform.

The invention relates to a training method of a verification model, a sharing method of the verification model, a system and a medium. The method according to the embodiment comprises the steps: constructing a consistency detection model comprising feature layers for carrying out different detection tasks; and training a consistency detection model through sample data; the constructed model can determine whether different text data are consistent or not from multiple levels; whether addresses filled by different people are consistent or not can be rapidly determined through the model, the working efficiency is improved, meanwhile, a model sharing method is provided, under the condition that source data are not shared, the model is trained through more source data, the model can adapt to wider scenes, and the data safety is improved.

The invention discloses a sharing method of warning data in a province and region integrated scheduling automation system. The method comprises the following steps of: 1) after a province scheduling automation system warns a sharing client to receive a station sharing request sent by an on-demand model sharing tool, positioning a real-time warning sharing service of a service end of a region scheduling automation system through a service bus and sending a warning sharing request to a warning sharing service end of the region scheduling automation system; 2) maintaining a real-time warning array by the warning sharing service end of the region scheduling automation system and receiving and preserving all warnings; sending real-time warning data and response reports of the requested station in request information to a warning sharing client of the province scheduling automation system; and 3) after the warning sharing client receives the warning data sent which is sent by the service end, analyzing the data and then supplying the analyzed data to a warning window program to use. According to the steps of the sharing method provided by the invention, the real-time warning information of the station at a region scheduling side can be seen by the province scheduling in the first place.

The invention discloses a category analysis method-based near-infrared spectrum material content indexing method, which comprises the following steps of an instrument portrait building step, namely obtaining a feature information through an instrument portrait, dividing the feature information into multiple tags and obtaining tag weight of each tag; a content indexing step, namely creating a content indexing table through a category analysis method to obtain a classification mode of the component content and obtaining a content prediction value of the component content of the substance according to the classification mode of the component content and the corresponding tag weight; and a gradient prediction step, building a gradient prediction value according to the content prediction value and pushing the gradient prediction value to a user. The invention further discloses electronic equipment for executing the method and a computer readable storage medium of storing the method. Regression analysis is replaced with a category analysis method, so that the problem of an interstation difference is effectively solved, and model sharing between similar instruments is achieved under the condition of not using model transfer.

The invention relates to the technical field of engineering project informatization management application, and particularly discloses a method for automatically converting and releasing a model, which comprises the following steps of: a program is converted through a command line provided by model conversion target software in combination with an automatic model conversion service client for comprehensive scheduling in the patent; a lightweight model system, a main system and a user side are subjected to comprehensive scheduling, a model automatic conversion service client automatically downloads a model file in the main system, after model conversion, the model file is uploaded to the lightweight model system to generate a model sharing address, and then the model automatic conversion service client feeds back the generated model sharing address to the main system; the user side with the access authority can inquire the model file after logging in the main system, and the user only needs to upload the model instead of comprehensively understanding the complex process of model conversion and carrying out manual operation in the past, so that the training cost of service system use, the user operation complexity and the operation time cost are greatly reduced.