18 results about "Problem domain" patented technology

Methods and systems are disclosed for solving a regression problem, for example a data regression and / or a differential equation problem, over a problem domain. The method comprises: receiving or determining, by a classical computer, a regression problem description and a set of kernel points in the problem domain; receiving or determining, by the classical computer, a trial function associated with the regression problem, the trial function being based on a quantum kernel and being parameterized by kernel coefficient(s); determining, using a quantum computer, for each of the kernel points, a kernel value of the quantum kernel and / or a kernel derivative value of a derivative of the quantum kernel; determining, by the classical computer, a set of optimal kernel coefficients based on the kernel value and / or kernel derivative value and determining, by the classical computer, a solution function based on the trial function and the set of optimal kernel coefficients.

A method is provided for watermarking a machine learning model. A sequence of bits is generated. The sequence of bits may be text characters divided into chunks. A selected plurality of input samples from training data is divided into subsets of input samples. All of the input samples of each subset of the subsets of input samples are labeled with a same first label in a problem domain of the ML model. Each chunk is combined with a subset of the labeled subsets to produce a plurality of labeled trigger samples. Each trigger sample of each set of the plurality of sets is relabeled to have a second label different from the first label and in the problem domain to produce a relabeled set of trigger samples. The ML model is trained with the training data and the relabeled trigger samples to produce a watermarked ML model.

An integrated method for supporting complex system design and optimization selects first graphical-textual model frameworks in a problem domain; implements functional requirement verification in a validation domain and guides a physical architecture model; selects second graphical-textual model frameworks in a solution domain to establish a physical architecture model; generates an integrated simulation model oriented towards functional and non-functional requirements in the validation domain; selects third graphical-textual model frameworks in an optimization domain to establish a multidisciplinary optimization model, and establishes binding with preset subsystem parameters in the physical architecture model of the solution domain and output values of variables related to an objective function; automatically feeds an optimal optimization scheme obtained through iterative solving based on intelligent optimization algorithms to a system design model for automatic verification. The integrated method breaks down barriers between models of various stages, improves the development efficiency, and avoids the difficulty of seamless integration.

A method for determining a quantum neural network (QNN) configuration for classification in a problem domain by generating a population of candidate QNN configurations based on a number of feature val

The application discloses a model-based deep space exploration communication constellation system design method, and belongs to the field of spacecraft digitization technology.The method comprises the following steps: constructing a unified digital model as a single data source throughout the whole life cycle; capturing stakeholders and their interaction interfaces through business scenario analysis, and obtaining top-level service indicators by decomposition to construct a problem domain model; taking the problem domain model as input driving, performing white-box analysis on system capability and architecture, generating a preliminary design scheme containing key parameters, and associating the key parameters to the top-level service indicators; taking the preliminary design scheme as input, performing refined design on system behavior and physical architecture; integrating professional simulation tools to perform computable verification on the key parameters, and automatically updating the corresponding parameters in the preliminary design scheme; and automatically generating a satellite development task book based on the iteratively converged design scheme.The application realizes full-chain digital forward design of a deep space exploration constellation system from demand to satellite.

This application provides a method and apparatus for evaluating comprehension and innovation inspiration capabilities based on multi-person discussion texts, relating to the field of comprehension and innovation inspiration capability evaluation technology. The method includes: acquiring multiple dialogue texts obtained from multi-person discussions, and constructing an influence graph representing the problem domain and solution domain based on the multiple dialogue texts; one dialogue text per participant; employing a heuristic method, utilizing a natural language model to process the multiple dialogue texts, evaluating the comprehension ability of each participant, and evaluating each participant's problem-solving ability and creative thinking based on the influence graph. The method and apparatus provided in this application for evaluating comprehension and innovation inspiration capabilities based on multi-person discussion texts are used to comprehensively and accurately evaluate the comprehension and innovation inspiration capabilities of each participant in a multi-person discussion scenario.

The invention relates to the technical field of electric digital data processing, in particular to an AI large model calling load balancing method, device and equipment and a medium. The method comprises the following steps: determining a skilled field label of each alternative AI large model, and screening out a preset problem field category matched with the skilled field label; when a to-be-processed problem request set is detected, screening out a problem domain category of each problem request through a problem request classification model; grouping the problem requests to obtain at least one problem request group; and distributing each problem request group to each alternative AI large model for processing according to the skilled field label of each alternative AI large model and the problem field category of each problem request group. According to the embodiment of the invention, after a batch of problem requests needing to be processed are obtained each time, each problem request can be shunted to a plurality of AI large models which can be called to be processed, so that the efficiency and the stability of the AI large model calling process are improved.

This invention discloses a parallel contact detection method for FDEM based on dual-domain activation, comprising: constructing a hash domain and a contact domain; dividing the problem domain into hash units; marking hash units containing TET4 units as active states; and constructing a contact domain centered on the host unit according to spatial neighborhood rules; sorting the TET4 units according to hash keys; traversing the sorted sequence to count the number of units in each hash unit, generating an index interval, marking the active states, and constructing an active unit list; based on the active unit list, sequentially selecting the host hash units, determining the visitor hash unit set according to spatial neighborhood rules, and performing pair detection on the TET4 units in the host and visitor units after activation state filtering; and generating a global contact pair set through two rounds of scanning for subsequent contact mechanics calculations. This invention, through a dual-domain activation mechanism and parallel sorting using GPU, limits the contact search to local active regions, avoiding global traversal and dynamic hash reconstruction.

The present invention provides a method and application for simulating a large deformation process of an impact based on a meshless interpolation technique, which belongs to the field of computational mechanics and numerical simulation technology, including model parameter setting, discretization of the problem domain and the time domain, iterative updating of the variables to be determined in the problem domain using the regularized Newton method, solving the local deformation gradient at the next moment, solving the node acceleration, updating the node coordinates, updating the material point coordinates and physical information, completing the simulation of the impact process, and the like. The present invention uses the regularized Newton method for the objective function of the local maximum entropy, and by constructing a simple shape near the calculation point, provides a method for quickly, accurately, and efficiently determining the initial point, thereby ensuring the solution efficiency, and using the Nelder‑Mead algorithm to solve the extreme case iteration point, thereby ensuring the stability of the solution, and is used for the complete simulation of the large deformation process of materials under ultra-high-speed impact, and has the characteristics of high calculation accuracy and strong stability.

Methods, computing systems, and computer-readable media for robust classification using active learning and domain knowledge are disclosed. In embodiments described herein, global feature data (such as a list of keywords) is generated for use in a classification task (such as a NLP text classification task). Expert knowledge, based on decisions made by human users, is combined with existing domain knowledge, which may be derived from existing trained classification models in the problem domain, such as keyword models trained using various datasets. By combining the expert knowledge with the domain knowledge, global feature data may be generated that is more effective in performing the classification task than either a classifier using the expert knowledge or a classifier using the domain knowledge.

The algorithm development process begins by receiving ontology information defining aspects of a problem domain, the ontology information including a list of types and associated attributes. This information is then transformed into homological descriptions. A chain complex is created based on mathematical postulates, wherein the chain complex satisfies predefined homomorphism rules. The system computes a basis for the chain complex. The basis is partitioned into an input set and an instruction set to define a synthesized algorithm. The synthesized algorithm is added to an algorithm knowledge base. Finally, the synthesized algorithm is executed to solve a problem in the problem domain.

The present application relates to a kind of simulation software skin particle generation method suitable for complex geometry, belong to computer simulation and industrial software technical field.This method is first according to the geometric description of given boundary, generates boundary particle in problem domain.According to this, the whole problem domain space is traversed, and the connectivity between all particles is judged by applying flood filling method, and the particles of different connected regions are obtained.The flood filling method can be implemented by non-recursive version to avoid overflow of recursive stack space caused by large scale of particles.This method is especially suitable for large-scale particle generation in complex geometry of high resolution multi-connected body, which guarantees the needs of high-precision simulation of different multi-physical body and multiple phase in complex scene for simulation software.

A method is provided for watermarking a machine learning model. A sequence of bits is generated. The sequence of bits may be text characters divided into chunks. A selected plurality of input samples from training data is divided into subsets of input samples. All of the input samples of each subset of the subsets of input samples are labeled with a same first label in a problem domain of the ML model. Each chunk is combined with a subset of the labeled subsets to produce a plurality of labeled trigger samples. Each trigger sample of each set of the plurality of sets is relabeled to have a second label different from the first label and in the problem domain to produce a relabeled set of trigger samples. The ML model is trained with the training data and the relabeled trigger samples to produce a watermarked ML model.

Provided is an operation method of a security requirement recommendation system including inputting an attack scenario of an advanced persistent threat (APT); and estimating a specific APT attack case similar to the attack scenario based on a case-based problem domain ontology including characteristic models of the APT attack cases, and recommending a security requirement corresponding to the specific APT attack case.

The invention discloses a federated agent-based distributed problem solving system and an excitation method thereof, and belongs to the technical field of artificial intelligence, and the system comprises a plurality of participant nodes, each node is provided with one or more specially-trained large language models, and the participant nodes are used for processing a specific problem domain; the central server is equipped with a high-performance large language model and is responsible for coordinating each participant node and integrating and optimizing answers returned from each node; through an intelligent distribution algorithm, model independence guarantee, dynamic learning and optimization and incentive mechanism design, the problem solving efficiency and quality are improved, and a continuously advanced and self-optimized ecological system is promoted through an incentive mechanism, so that each participant can effectively improve the efficiency and quality of problem solving while protecting own interests. And the optimization and progress of the whole system are jointly promoted. The mode is particularly valuable for modern enterprises or organizations needing to deal with a large number of complex and specialized problems, and has extremely high practicability and popularization value.

The invention discloses an integrated method for supporting complex system design and optimization, which is applied to the technical field of system modeling, and comprises the following steps: selecting a first image-text model framework in a problem domain; functional requirement verification is achieved in the verification domain, and the physical architecture model is guided; selecting a second image-text model framework in the solution domain to establish a physical architecture model of the system; generating an integrated simulation model oriented to functional requirements and non-functional requirements in the verification domain; selecting a third image-text model framework in the optimization domain to establish a multidisciplinary optimization model, and establishing binding with preset subsystem parameters in a physical architecture model in the solution domain and output values of target function related variables; and an optimal optimization scheme obtained based on intelligent optimization algorithm iteration solution is automatically fed back to a system design model for automatic verification through an optimization domain. The integration method breaks the barriers among the models in all stages, improves the efficiency of the research and development process of a complex system, and avoids the problem that system optimization is difficult to be seamlessly integrated with system design and even simulation.

In a method of responding based on sentence paraphrase recognition for dialog system, main keywords of a domain and patterns thereof are recognized and extracted based on a morpheme analysis result in a pre-processing process. Question domains / sub-categories / dialogue-acts are classified using the morpheme analysis result and the extracted main keywords and patterns. Learning a model is performed using classification features of the classified question domains, sub-categories, and dialogue-acts as semantic features of query sentences, and sentence semantic vectors are extracted by measuring similarity between the vectors. A language model of letters is trained and similarity in expression and structure is analyzed by extracting a sentence expression vector based on the letter. An answer to a similar question is provided by generating a vector containing semantic and expressive information about an input query sentence based on analyzed semantic and expressive similarities, and finding a similar query sentence from FAQ knowledge using a paraphrase recognition technology. In a dialog system for automatic Q & A service such as a chatbot for customer consultation, it is possible to provide related answers by exploring question-and-answer knowledge (questions) that have similar meanings and intentions of input sentences (query) through paraphrase recognition technology.

A method for watermarking a machine learning model is provided. A bit sequence is generated. The bit sequence may be a text character divided into a number of tiles. A selected plurality of input samples from training data are divided into input sample subsets. All of the input samples in each of the subsets of input samples are tagged with a same first tag in a problem domain of an ML model. Each tile is combined with a subset of the marked subsets to produce a plurality of marked trigger samples. Each trigger sample of each of the plurality of sets is re-tagged to have a second tag different from the first tag and in the problem domain to produce a re-tagged set of trigger samples. The ML model is trained with the training data and the re-marked trigger sample to produce a watermarked ML model.