61 results about "Base class" patented technology

The invention discloses a personalized federal incremental learning method based on instance-level prompt generation, which realizes finer-grained feature capture of data by deploying a dynamic adjustable prompt generation module for a basic network of each client. The core of the module is to mine hidden attributes in data, and the attributes not only can significantly improve the generalization ability of the model for new classes, but also can realize effective migration of base class knowledge to the new classes. The hidden attribute features learned from the base class can be naturally generalized into unseen classes, so that the adaptability of the model to unknown distribution is enhanced while historical knowledge is kept. According to the method, the bottleneck of the static prompt pool in the aspects of expandability and adaptability is broken through, and a more flexible solution is provided for personalized federal continuous learning.

In a machine learning apparatus that learns data of a novel class with a smaller number of samples than data of a base class by continual learning, a feature extraction unit is pre-trained using first data and second data of the base class. The feature extraction unit receives an input of the data of the novel class to output a feature vector of the data of the novel class. A weight calculation unit calculates a classification weight of the novel class based on the feature vector. A graph model receives an input of the classification weight calculated and classification weights of all classes previously learned and outputs reconstructed classification weights. The graph model is trained by pseudo continual learning using third data of the base class. The first data, the second data, and the third data are different data.

The invention discloses a program decompilation prevention method and device and a medium, and the method comprises the steps: determining a target code required to be protected in a program source code, carrying out the obfuscation processing of the target code and a mixed factor code of the target code according to the obfuscation logic of the target code, generating an intermediate code, and carrying out the obfuscation processing of the mixed factor code of the target code through an SM4 symmetric encryption algorithm, the method comprises the following steps of: firstly, generating an intermediate code, encrypting the intermediate code to generate an encrypted Class file, then, carrying out hierarchical encryption on a preset loading class to generate a multi-layer encrypted loading class, and finally, packaging the encrypted Class file, the multi-layer encrypted loading class and a basic class to generate an encrypted JAR packet. According to the method, the Class file is encrypted through the code obfuscation technology and the SM4 cryptographic algorithm, layer-by-layer encryption is carried out on the self-defined loading class, the complexity after code obfuscation is greatly improved, and the security of source code software is improved.

The invention discloses an MRI (Magnetic Resonance Imaging) sequence generation system and method, and the system comprises a basic sequence generator module which is used for defining an abstract base class and packaging general logic in a sequence generation process; the independent development tool chain module is used for providing a development environment independent of a system framework and compiling a sequence generator source code into a binary module; the sequence registering and loading module is used for automatically discovering and loading the binary module and registering a sequence generator class inheriting the self-abstract base class through a reflection mechanism; the RESTful API interface module is used for receiving an external request and calling a corresponding sequence generator object according to a sequence type; the asynchronous task management module is used for managing life cycles and state tracking of asynchronous tasks; and the parameter configuration management module is used for centrally managing sequence parameter configuration. According to the system and the method, the development threshold is reduced, and the expansibility, the maintainability and the concurrent processing capability of the system are improved.

The application relates to the technical field of code data, and discloses a method for solving nested object dictionary and automatically modifying existing codes, which comprises the following steps: S1, a command line module reads and stores parameters input by a user, and obtains existing codes and related information; S2, a file adding module adds a base class file in a base class file module to root directory information of the existing codes, wherein the base class file module comprises a to_json method and a to_dict method; S3, a file modifying module transmits init code and serialization style to an attribute acquisition module; and S4, the attribute acquisition module returns an attribute dictionary to the file modifying module. By implementing a base class, all objects inherit the class, and expression and serialization of the objects are easily realized.

The application relates to a function module operation method, device and equipment and a storage medium, the method comprising the following steps: a first function module class corresponding to a first function module is created; the first function module class inherits a base class provided by a preset function module development framework; a data service interface for data interaction between modules and an execution interface for executing business logic in a module are arranged in the base class; first output data corresponding to a second function module relied on by the first function module is acquired through the data service interface in the first function module class; the first output data is data provided externally to the second function module through the data service interface in the base class; the execution interface in the first function module class is called, the first function module is operated based on the first output data according to preset business logic. The application reduces the complexity of control logic, reduces resource consumption and improves the decoupling degree between modules through a data driving mode.

In a machine learning apparatus of the present invention, a neural network outputs a base class classification and a novel class classification. A loss calculation part calculates losses in the base class and novel class classification. An updating part updates a weight based on the losses in the base class and novel class classification. The updating part updates the weight by providing the weight with a regularization term and a sum of the losses.

This invention provides a machine learning apparatus, method, and program that improve the classification performance of a classifier by utilizing local features of the input data. [Solution] A machine learning device that continuously learns a small number of new classes compared to the base classes, wherein the feature extraction unit 10 extracts features of the input data using weights learned based on the divided features of the input data. The base class classification unit 20 classifies the base classes based on the features of the input data. The new class classification unit 40 classifies the new classes based on the features of the input data. The attention attractor unit 30 regularizes the weights of the base classes and the weights of the new classes.

The invention discloses a numerical simulation program design method based on a microkernel architecture and related equipment, and relates to the technical field of numerical program development, and the key points of the technical scheme are as follows: packaging at least one basic component which is not related to a numerical simulation service in a nuclear reactor numerical simulation program to obtain at least one base class; constructing a microkernel based on the base class; setting at least one service expansion point for a corresponding base class in the microkernel according to business logic in different numerical simulation businesses, and abstracting each service expansion point into a virtual function interface; configuring a plug-in for the virtual function interface; calling a plug-in registration macro function to register each plug-in, and storing registration information into a plug-in registry; parameters input by a user are read from the file blocks, and entity plug-in class objects are constructed in combination with registration information in the plug-in registry; and according to the entity plug-in class object, loading the plug-in to a base class corresponding to the microkernel for assembly, and forming numerical simulation programs corresponding to different numerical simulation services.

The application discloses a generalized small sample semantic segmentation method based on word vector embedding-information self-completion, comprising: a base class learning stage: training a backbone network, a semantic word vector embedding module and a base class classifier using base class label samples to generate a base class prototype; a new class updating stage: training a new class classifier using one or five new class label samples to generate a new class prototype; an evaluation stage: inputting the base class prototype obtained in the base class learning stage and the new class prototype obtained in the new class updating stage into a self-completion module, using the base class prototype and the new class prototype to generate pseudo labels through initial prediction of the query image by the self-completion module, then extracting a query prototype in the query image through the pseudo labels, calculating the cosine similarity between the query prototype and the query image to generate an auxiliary prediction; finally, weighting the auxiliary prediction and the initial prediction to obtain a final prediction result. The application increases the inter-class distance of the model and excavates more information, and further improves the segmentation performance of the model.

This invention discloses a few-sample class-variable audio classification method based on a prototype adaptive network. The steps are as follows: 1) Extracting log-Mel spectrum from input audio samples; 2) In the basic stage, training an initial model consisting of a feature extractor, a prototype classifier, and a class-adjustable prototype adaptive network using base class samples; 3) Extracting representations from the base class training samples, calculating the mean of each base class representation as a prototype, and updating the prototype classifier; 4) In the incremental stage, if the number of categories increases, inputting the representations of the incremental class training samples into the class-adjustable prototype adaptive network to obtain the prototypes of each incremental class and updating the class-adjustable prototype adaptive network; if the number of categories decreases, deleting prototypes of unnecessary categories; 5) Then updating the prototype classifier using the class-adjustable prototype adaptive network; 6) Obtaining the classification category of the test sample. This invention can adaptively expand or shrink the prototype classifier, thereby continuously learning new categories while remembering or deleting old categories, achieving few-sample class-variable audio classification.

The application provides a method for adapting functions with various parameters in a virtual machine, comprising: creating a base class to store different virtual machine methods; creating a template class to provide a template function and containing various parameter function data, and returning an object array as a result to realize function calling of any virtual machine method; and creating an auxiliary class to encapsulate the parameter data, convert any parameter data into an interface of one data, realize adaptive calling of the function calling function with various parameters, store the function data with various parameters through the template function, and store any number of parameters into one data, so that any method in the virtual machine can be adapted, and any number and type of parameters can be adapted.

The application provides a visual-language model prompting method and system based on a coupling prompt field, and belongs to the fields of artificial intelligence and computer vision. The method comprises the following steps: mapping a base class task and a new class task to a shared feature space, defining a coupling prompt field, and enabling the base class and the new class task to form mutual constraints in the shared feature space; performing scale constraints on the coupling prompt field through projection layer norm alignment and coding layer layer-by-layer norm alignment; integrating the projection layer norm alignment loss and the coding layer layer-by-layer norm alignment loss, fusing the alignment loss and a task loss of a visual-language model, forming a model overall training target, and performing training; performing reasoning on the base class task and the new class task based on the trained model, and outputting a classification prediction result. The application solves the problems of existing end-to-end, decoupling prompt learning, base class-new class isolated optimization, easy falling into local optimum, norm drift and entanglement collapse, and improves the cross-task generalization and robustness of the model.

This invention discloses a few-sample incremental action recognition method and apparatus with test-time adaptive capability, relating to the field of video action recognition technology. The invention is based on a two-stage structure of "base class learning—new class adaptation": in the base class stage, the model learns general spatiotemporal features through sufficient samples and introduces a multi-head self-attention mechanism to improve generalization ability; in the new class stage, a hierarchical test-time adaptive mechanism is used to achieve self-updating of the model during inference. Using this invention, not only is the model's responsiveness to new class samples improved, but catastrophic forgetting and bias accumulation caused by noisy samples are also effectively suppressed, thereby achieving a stable and efficient incremental learning process, and thus improving the accuracy and efficiency of few-sample action recognition.

According to the software test engineering conversion and test method and device, the storage medium and the computer equipment provided by the invention, when the engineering structure is converted, the primitive class in the original test engineering test class is replaced with the new base class; the new base class is obtained through reconstruction by adding a data source setting capability and a data reading capability on the basis of the primordium class, so that test data of a specific scene can be accurately read; then, an original test method containing a plurality of scene cases and a test logic core method of the original test method are obtained through analysis from the test class, the workload of generating an independent test method for each scene case can be greatly reduced by integrating universal test logic in the core method, and the independent test method not only has an independent execution entry, but also has a test logic function; and the corresponding test data line in the data storage file can be accurately read and executed through the new base class. Therefore, a new test project is obtained through structure conversion based on each independent test method, and a large amount of workload and potential errors caused by manual modification can be avoided.

The present invention relates to a few-shot semantic segmentation method based on a learner integration strategy. Aiming at the problem of base class bias existing in previous related work, that is, the base class target regions are easily misactivated, the present invention proposes to introduce an additional auxiliary branch (base learner) on the basis of a traditional model (meta-learner) to explicitly identify the base class targets, and then adaptively integrate the rough prediction results of the two learners to generate accurate segmentation masks. Considering the sensitivity of the meta-learner, the present invention further introduces an adjustment factor to estimate the differences between input images to promote model integration inference. Generally speaking, the present invention provides a novel and general solution for improving the problem of misactivation of base class targets by existing few-shot segmentation models, which is not limited to a certain base learner and meta-learner, significantly improves the performance of existing segmentation networks and can be extended to more practical and challenging generalized settings, showing good application prospects and potential.

The embodiment of the invention provides a component adaptation method and device, computer equipment, a computer readable storage medium and a computer program product, and belongs to the field of software. The component adaptation method comprises the following steps: acquiring an interface of a source framework component and a predefined component base class, wherein the component base class comprises a universal interface mapping rule; mapping an interface of the source frame component into an interface of a universal component based on the universal interface mapping rule, so as to convert the source frame component into the universal component; wherein the interface of the universal component is determined according to the interfaces of various framework components, and the universal component is suitable for various frameworks. According to the technical scheme of the embodiment of the invention, a method for efficiently converting the specific framework component into the universal component can be provided under the condition of not invading framework type selection, and the universal component is suitable for multi-framework access, so that the design cost of a component provider and the access cost of an access party are reduced, and the development efficiency is improved.

In a machine learning apparatus that learns data of a novel class with a smaller number of samples than data of a base class by continual learning, a feature extraction unit is pre-trained using the data of the base class. The feature extraction unit receives an input of the data of the novel class to output a feature vector of the data of the novel class. A weight calculation unit calculates a classification weight of the novel class based on the feature vector. A graph model receives an input of the classification weight of the novel class and classification weights of all classes previously learned to output reconstructed classification weights. The graph model is trained by pseudo continual learning using alternative data of the base class to learn a dependency between the base class and the novel class by meta learning.

The embodiment of the invention provides a system operation and maintenance method and device, computer equipment and a computer readable storage medium, and the method comprises the steps: obtaining base class interface data of a system adapter and business logic data of a target manufacturing execution system in response to an access request for the target manufacturing execution system; according to the base class interface data and the business logic data, constructing derived class interface data corresponding to the target manufacturing execution system; and updating the derived interface data into a system adapter, wherein the updated system adapter is used for performing data conversion and communication between the automatic test equipment and the target manufacturing execution system through the derived interface data corresponding to the target manufacturing execution system. According to the embodiment of the invention, the code development efficiency in a system operation and maintenance scene can be improved, and the system operation and maintenance cost is reduced.

The present invention discloses a method and a storage medium for improving the readability of register model modeling code. A method for improving the readability of register model modeling code includes the following steps: creating a UVM register model, deriving from the field base class of the UVM register model to generate a subclass; expanding or adding a series of interface methods in the subclass to support method calls of enumerated type values; creating an enumerated type variable corresponding to the register field of the preset UVM register model; using the subclass to replace the field base class to reconstruct the example register; and using the interface method of the subclass to perform read and write access to the register field. The present invention can improve the coding method for the readability of register fields, avoid repeatedly referring to the design documents, save time for design and verification developers, and improve their work efficiency.

Continual learning techniques are described for extending the capabilities of a base model, which is trained to predict a set of existing or base classes, to generate a target model that is capable of making predictions for both the existing or base classes and additionally for making predictions for new or custom classes. The techniques described herein enable the target model to be trained such that the model can make predictions involving both base classes and custom classes with high levels of accuracy.

The application discloses a development framework and method for asynchronous remote calling based on Qt, the development framework for asynchronous remote calling based on Qt comprises a base class and a public calling function callFunc, the base class realizes a callSelfFunc function, a function is found and called through a reflection mechanism provided by Qt, and a called module inherits the base class, the public calling function "callFunc finds a called module object according to a module name, generates a QEventLoop object, and calls the callSelfFunc function of the called module through an invokeMethod provided by Qt, the development framework for asynchronous remote calling based on Qt is based on a development framework for message transmission based on Qt, a return value of the called module is called through the development framework for message transmission and is synchronously returned to a calling module, the logical splitting problem is solved, the implementation complexity is reduced, meanwhile, a non-thread-safe "called function can be called in a multi-thread environment, and some special application scenarios are adapted.

The invention discloses a small sample target detection method based on a decoupling mechanism and metric learning, and the method comprises the steps: 1, dividing a training set into a base class data set and a new class data set according to classes, and enabling the base class data set to be used for pre-training to construct a base model containing abundant universal features, the new class data set is a group of small sample data sets with different visible image numbers generated through sampling and is used for simulating a sample scarce scene; 2, training on the base class data set to obtain a base model; and 3, finely adjusting the base model on the new class data set. According to the method, an adaptation enhancement module is provided based on a decoupling mechanism, features are projected to downstream task subspaces, and multi-task decoupling is achieved; in the back propagation stage, decoupling of the class-related features and the class-independent features is achieved; a mixed category characterization measurement module combining a single-mode positive category characterization vector and a multi-mode negative category characterization vector is provided, the intra-category feature variance is reduced, and the accuracy of target detection is improved.

Providing adapting a base classifier to novel classes. A system (100) for adapting a base classifier to one or more novel classes is disclosed. The base classifier classifies instances into base classes by extracting feature representations from the instances using a feature extractor and matching them to class representations of the base classes. The base classifier is adapted using training data for the novel classes. Class representations of the novel classes are determined based on feature representations of instances of the novel classes. The class representations of the novel classes and the base classes are then adapted, wherein at least one class representation of the novel classes is adapted based on class representations of the base classes and at least one class representation of the base classes is adapted based on class representations of the novel classes. The adapted class representations of the base classes and the novel classes are associated with the base classifier.

The invention discloses a class increment semantic segmentation method based on feature purification gradient projection, which is characterized in that a model is trained to learn the optimal segmentation capability of a current task, and then feature purification and construction of a specific feature space of the task are carried out; the method specifically comprises the steps of base class model training, feature purification, task specific subspace construction, gradient projection optimization and the like. Compared with the prior art, the method has the advantages that dependence on complete historical data is reduced under an incremental learning framework, the model does not need to retrain all categories from zero, the storage and calculation cost is greatly reduced, meanwhile, the training efficiency and the resource utilization rate are improved, and the method is suitable for popularization and application. Large-scale class increment semantic segmentation can be efficiently realized under the condition of finite computing power, and a more economical solution is provided for model deployment in the fields of automatic driving, intelligent monitoring and the like.

An embodiment of the present invention provides a class loading protection method and device based on white rules. Among them, the method includes: obtaining the stack of the target thread through the first monitoring point, scanning the stack to determine whether the target thread has an operation of loading a class; if there is an operation of loading a class, calculating the bytecode feature code of the file corresponding to the file path through the file path corresponding to the target thread provided by the first monitoring point, and matching the bytecode feature code with the whitelist to obtain a first matching result, where the whitelist includes the bytecode feature code and the class name of the legally loaded class; obtaining the class name of the class loaded by the target thread through the second monitoring point, and matching the class name with the whitelist to obtain a second matching result; if the first matching result and / or the second matching result is a failure to match, the target thread is a thread of an attack script, and an alarm message is generated; otherwise, the operation of loading the class is continued. This method realizes effective protection against unknown vulnerabilities.

The application provides an annotation and base class-based data screening method and system, comprising: establishing a programming language base class; establishing an aspect module and performing corresponding business logic assignment processing on fields according to the programming language base class and actual business requirements; assembling base class fields in the programming language base class into structured query language query conditions through the established Mybatis interception module, and then executing the structured query language to obtain corresponding results. The application puts all data screening related fields into a certain base class and extracts all assignment and business logic processing parts of the fields to an aspect class based on the industry mainstream open source framework SpringBoot, and the aspect class uniformly processes, so that the project group screening function code sharing is realized, the code is non-invasive, the coding speed of the developer is greatly improved, and the bug repair speed is also improved.

This invention relates to the field of computer technology, and more particularly to a method, apparatus, computer device, and storage medium for managing pop-up views. Based on a received trigger event of a pop-up view, the corresponding pop-up component is invoked to generate the corresponding pop-up view. An interface utility class file for managing the pop-up views is obtained based on a pre-established base class inherited from the pop-up view, and the pop-up views are managed based on this interface utility class file. In this invention, by inheriting the base class of the pop-up view, only one set of code needs to be written and maintained for the same pop-up generation and disappearance, providing methods for business use. Subsequent adjustments or modifications only require maintenance within this method, simplifying the maintenance of the pop-up view code and improving the management efficiency of pop-up views.

The embodiment of the application discloses a dynamic entity input method and device based on a low-code development platform, a storage medium and a terminal device, and relates to the field of data processing. The embodiment of the application enhances the flexibility and adaptability of the data model by constructing a base class model and its attribute information and creating a subclass model, effectively responds to changes in business scenario requirements, reduces repeated configuration, and improves modeling efficiency. With the help of a scene and a reference operation template configuration interface, multiple subclass models can be easily created and managed on an interface for operation, the subclass model is bound to a specific scene, and intuitive display and interaction of data are realized. The entity input interface is bound to the model attribute, the dynamic and intelligent data input during the running period is realized, the user can display the entity input interface by triggering the attribute input box, and select and input data. After data input, the corresponding input box is automatically filled in, the work efficiency is improved, and the data accuracy is ensured.