16 results about "Turing machine" patented technology

The invention discloses a multi-stage task processing method and system based on an intelligent Agent model, and relates to the technical field of task processing, and the method comprises the steps: collecting heterogeneous data streams through a distributed sensor network, and generating a dynamic feature vector through a feature encoder inspired by quantum annealing; calculating a path expected utility value by using a Bayesian optimization algorithm, and projecting a high-order task space to a Kupman space; starting multi-thread asynchronous calculation, collecting execution state data in real time and constructing a causal graph model; the short-term execution logs are integrated through a neural Turing machine, edge computing nodes are called for distributed knowledge extraction, a multi-mode interpretable report is generated, and a long-term memory library is updated. According to the method, by starting multi-thread asynchronous calculation, the execution state data are collected in real time, the causal graph model is constructed, the execution result matrix with the confidence score is output, and the stability and reliability of task execution are improved.

A computer-implemented method, according to one approach, includes: causing a multi-layer neural network to evaluate a user query received from an endpoint device by processing the user query with one or more initial layers of the neural network. In response to receiving an output from the initial layers at a supplemental threat detection layer of the neural network, the user query processed with the threat detection layer. Processing the user query with the threat detection layer includes using a classifier to compare the user query to activation data outlining known illegitimate queries. Moreover, combinational reasoning is used to determine whether the user query is legitimate based at least in part on: an output of the classifier, and behavioral information received from a threat intel pattern. The user query is further intentionally rejected in response to the threat detection layer determining the user query is not legitimate.

The invention relates to the technical field of data processing, in particular to a large model reasoning optimization method based on context incremental updating, which comprises the following steps: processing a multi-modal data stream through timestamp alignment and a filtering algorithm, extracting features by adopting a shared encoder and a private encoder, and realizing feature decoupling through a depth information bottleneck principle. A dynamic emotion map is constructed by using Gaussian process regression and a random process algorithm, and self-adaptive updating control is realized by combining meta-learning and Bayesian optimization. Incremental state management is realized by adopting a neural Turing machine, reasoning consistency is guaranteed through a generative adversarial network, model parameters are optimized in combination with a digital twin system and reinforcement learning, and a mental health service response is finally generated through a conditional generation model and hierarchical reinforcement learning. According to the method, the problem of asynchronism of multi-modal emotion feature dynamic evolution and context increment updating is effectively solved, accumulated drift of emotion state tracking is eliminated, and the continuity of reasoning logic is guaranteed.

The invention discloses an intelligent manufacturing production scheduling optimization method based on artificial intelligence. The method comprises the following steps: S1, collecting and preprocessing production data; s2, modeling by using a neural Turing machine, storing and updating a scheduling history, and generating a task scheduling sequence; s3, task historical data are stored through the scene memory network, and a scheduling strategy is generated; s4, constructing an intelligent prediction model based on the scheduling sequence and the strategy, predicting a task execution sequence, a resource demand and an equipment utilization rate, and forming an optimization suggestion; s5, optimizing the intelligent prediction model by adopting reinforcement learning, and adjusting a task scheduling sequence and resource allocation; and S6, based on the optimized intelligent prediction model and the scheduling strategy, updating a scheduling sequence in real time, and dynamically configuring resources. Through the neural Turing machine, the scene memory network and reinforcement learning, production scheduling and resource allocation are optimized, efficiency is improved, waste is reduced, and production changes are flexibly coped with.

A computer-implemented method, according to one approach, includes: causing a multi-layer neural network to evaluate a user query received from an endpoint device by processing the user query with one or more initial layers of the neural network. In response to receiving an output from the initial layers at a supplemental threat detection layer of the neural network, the user query processed with the threat detection layer. Processing the user query with the threat detection layer includes using a classifier to compare the user query to activation data outlining known illegitimate queries. Moreover, combinational reasoning is used to determine whether the user query is legitimate based at least in part on: an output of the classifier, and behavioral information received from a threat intel pattern. The user query is further intentionally rejected in response to the threat detection layer determining the user query is not legitimate.

The invention provides an intelligent question and answer method and system, and the method comprises the steps: carrying out the semantic search of a user question, obtaining a semantic search result, carrying out the preprocessing of the semantic search result and a PDF text collected by a system, obtaining target data, and finally, carrying out the processing of the target data. Processing the target data by utilizing an intelligent energy Turing machine combining algorithm control and intelligent energy collection to obtain reply information aiming at the user question; the purpose of quickly and accurately answering the questions of the user is achieved, and the user experience is greatly improved.

The invention discloses a digital teacher knowledge modeling and content generation method based on artificial intelligence. The method comprises the following steps: constructing a teaching knowledge graph and forming a capability graph; generating a node vector, an edge vector and a path mode set; inputting the node vector, the edge vector and the path mode set into a Hopfield network, and executing pre-training; outputting a corresponding teaching knowledge path sequence and an association weight thereof based on teaching task information extracted from the teaching knowledge graph and the capability graph; generating a teaching content organization plan; according to the teaching content organization plan, generating structured teaching content including teaching explanation text, teaching example text, interactive practice tasks and evaluation question content; and performing increment adjustment on the Hopfield network memory state and the external memory of the neural Turing machine. The method integrates knowledge modeling, intelligent memory and content generation capabilities, and is suitable for intelligent education scenes.

The application relates to a calculation method and system based on constrained network minimum action evolution, belonging to the cross field of computer science and computational physics. The technical problem to be solved by the application is to overcome the limitation that the existing calculation paradigm completely depends on instruction sequence execution. The technical solution of the application is as follows: receiving constraint declaration information input by a user, including calculation element declaration, partial order relation declaration and action contribution item declaration; constructing an initial coupling network and initializing coupling strength according to a chaotic rule; under the constraint of the partial order relation, optimizing the coupling network through an iterative search method, searching for a target network configuration which makes the total action approach minimization; and performing coarse-grained observation on the target network configuration and outputting an emergent result. The application breaks through the existing Turing machine calculation paradigm, for the first time, takes the minimum action principle as the core driving force of calculation, replaces the instruction sequence with the partial order relation as the sequential constraint of a program, makes the program automatically evolve into a macro result from the constraint, realizes the fundamental innovation of the calculation paradigm, and can be applied to the fields of physical simulation, material design, cosmology research, new type of calculation hardware design and the like.

The invention relates to a large language model length generalization data driving method and device based on the thought of a Turing machine, and the method comprises the following steps: obtaining an input problem, and unfolding the problem into a thinking chain comprising a plurality of linear ordered steps through linear expansion; in the process of expanding the problem into the thinking chain, taking a basic solution generated in operand retrieval and reasoning steps and a logic control statement as atomic states of the steps; based on a thinking chain of an atomic state, reasoning is carried out by utilizing a large language model, before reasoning is carried out on each step, operands involved in the current step are obtained through operand retrieval in advance, and large language model length generalization data driving is realized. Compared with the prior art, the method has the advantages of realizing universal and effective length generalization data driving, eliminating potential shortcut learning, long-distance attention, reasoning action decoupling and the like.

The invention discloses a character processing method and device based on novel artificial intelligence, and relates to the technical field of character processing of artificial intelligence, the method can correct grammar and character errors of a user under the requirement of the user, provide sentence pattern rewriting suggestions, and can improve the character processing efficiency by understanding chapter writing intentions and summarizing the expression advantages of an original text. Defects and defects existing in character expression are pointed out, and corresponding improvement suggestions are given at the same time. And according to the intelligent energy Turing machine and corresponding improvement suggestions, carrying out defect checking and defect repairing on the optimized and rewritten text content so as to supplement missing key information of the optimized and rewritten text content, improve the rationality of a text structure and improve language coherence. By performing multiple rounds of correction and optimization on the text contents and expressions, a user is helped to more accurately and smoothly reflect deep context understanding of the text contents and expression intentions of chapters of the text in a text correction process.

The invention discloses a calculation method based on a spiking neural network, and the method comprises the steps: carrying out the discretization of the membrane potential of neurons in the spiking neural network SNN through employing a clustering method, dividing the membrane potential into a plurality of discrete states, and enabling each state to correspond to a clustering center; by controlling a clustering error, the finiteness of a state space is ensured, and the state space is equivalent to a state set of a target Turing machine; mapping a tape symbol of a target Turing machine into a discrete interval of a synaptic weight, and realizing read-write operation by dynamically updating the synaptic weight; a state transition function of a target Turing machine is simulated and realized through dynamic characteristics of neurons and synapses of SNN, simulation of a Turing machine structure is completed through the steps, and reading, writing, moving and state transition are simulated in a brain-like calculation system, so that general calculation is realized. According to the invention, the spiking neural network has general computing power besides image recognition and voice processing, so that the application of the spiking neural network is expanded.

The invention discloses a demand-driven Turing Machine (NDTM) model method, which is used for a maritime affair remote sensing detection task, and is used for solving the limitation of a current deep learning system in the aspects of autonomy, generalization and interpretability. According to the method, symbolic reasoning and state transition based on reinforcement learning are combined, so that an intelligent agent can internally represent dynamic requirements, select a self-adaptive strategy and execute an interpretable decision-making process. In order to verify the effectiveness, the method is applied to a marine remote sensing image ship detection task: a proper YOLOv11 sub-model is dynamically selected through a lightweight reasoning module according to deduced internal requirements instead of being embedded into a fixed detection framework. Experimental results on a standard remote sensing ship detection data set show that the NDTM-guided framework not only improves the detection precision and consistency, but also enhances the interpretability of the decision process. The research provides a generalized and explainable framework for the intelligent agent, and lays a foundation for autonomous decision-making in maritime affair application.

The present invention discloses a formal language and automaton simulation system based on the Android platform, comprising a user interaction layer, a data processing layer, an intermediate layer, and a storage layer based on the Android platform; the user interaction layer provides an automaton canvas for creating and modifying finite automata, pushdown automata, Turing machines, and regular expressions; the data processing layer is used for simulating finite automata, pushdown automata, Turing machines and extended Turing machines, and regular expressions and related data processing during automaton simulation; the intermediate layer is used to serialize automata drawn on the canvas into automaton files and deserialize automaton files; and the storage layer is used to manage automaton files. The present invention can realize the visual construction and simulation of various types of automata, the persistent storage of model data, and the processing of known errors, making it more suitable for use by beginners and teaching demonstrations.

The invention creates a system based on a block chain, and automatically records bookkeeping assets and legal behaviors by using an intelligent contract. The bookkeeping type assets represent property rights, and the governance rights are reflected as role rights in an economic organization. Law behaviors are defined as processes for executing the rights, and the state of the rights can be changed due to the right executing result. According to the system, a model is established by referring to a Turing machine, and the core is to define a structured data object, determine a state transition rule system and realize an algorithm and a process of state evolution. The three types of smart contracts are a register book contract used for defining and recording data objects, a behavior specification contract used for defining rules, and a manager contract used for verifying preconditions and realizing legal consequences. The model automatically controls legal behaviors by using an intelligent contract and records the legal behaviors on a block chain. In order to explain and explain the system, example codes and test scripts are specially included, and specific implementation methods of automatic management legal behaviors and related bookkeeping are displayed in detail.

The present invention discloses a scheduling method for the MES system in the lithium battery industry based on deep learning, which includes the following steps: S1, collecting and preprocessing lithium battery production data; S2, constructing a twin-tower Siamese network to encode the standardized sequence and generate a similarity matrix; S3, using Transformer-XL to model long-term dependencies and extract cross-cycle scheduling relationships; S4, constructing the scheduling relationships into an associated weight tensor, inputting it into a neural Turing machine, and generating an original scheduling instruction stream through read and write operations; S5, performing conflict detection and dependency rearrangement based on the controller memory comparison mechanism, and iteratively updating the state of the neural Turing machine; S6, feeding back candidate solutions to the MES execution layer and updating the model parameters. The present invention integrates multiple model structures, realizes intelligent modeling and dynamic optimization of lithium battery scheduling, and effectively improves the scheduling efficiency, response speed and system adaptability.

The present invention provides a Turing machine virtual simulation system based on Unity, which relates to the field of computer virtual simulation technology. This system models the Turing machine, designs the actions of the Turing machine for reading, writing, and state transition, uses the Turing machine to read and write characters and perform state transition to simulate five algorithms, and finally presents the process of the Turing machine simulating the algorithms in the form of 3D animations, and records the number of times the Turing machine reads and writes and the number of tape squares consumed to calculate the algorithm complexity. When using the Turing machine to simulate algorithms, different state transition equations are designed for different algorithms. At the same time, considering that the device performance and operating environment of different platforms are different, in order to improve the portability of use, the system is published in the form of WebGL based on WebGL technology. This system can not only utilize the execution process of simulating multiple classic algorithms, but also simulate the recursive function call process, realizing the function of Turing machine algorithm simulation.