287 results about "Neuron network" patented technology

The invention provides a photovoltaic station intelligent operation and maintenance simulation method and system based on digital twinning, and relates to the technical field of photovoltaic station operation and maintenance, and the method comprises the steps: constructing a digital twinning hierarchical structure through multi-modal detection signals through discrete wavelet packet transformation, tensor decomposition and a graph neural network, and determining structured data; extracting fault features by using a dynamic neural network, predicting a fault probability based on a Markov model, and triggering maintenance early warning; inputting the early warning signal into a maintenance strategy model based on multi-agent reinforcement learning, and generating a candidate maintenance time window set; the maintenance scheduling module optimizes sorting to obtain an optimal time window sequence, and the resource allocation module generates a personnel material allocation scheme by adopting a genetic algorithm; and iteratively optimizing the strategy parameters based on the maintenance cost, and outputting a final operation and maintenance implementation scheme.

The invention provides an intelligent material damage diagnosis method and system based on magnetic characteristic data fusion, and relates to the technical field of nondestructive testing, and the method comprises the following steps: collecting magnetic characteristic data by using a magnetic stress detection robot; performing multi-scale decomposition on the data to construct a magnetic domain evolution characteristic spectrum; inputting the characteristic spectrum into a composite neural network to extract a material damage characteristic matrix; constructing a fusion feature vector by using tensor dimensionality reduction and a multilayer belief transfer network; and realizing damage calculation through temperature field modulation and self-adaptive weight distribution to obtain the damage grade and the residual life. According to the invention, accurate diagnosis of the material damage in the welding seam area of the pressure vessel is realized.

A system for AI-supported cash flow management and forecasting, consisting of: a data acquisition unit configured to retrieve transaction data from multiple financial data sources, the data acquisition unit incorporating encrypted API connections and secure data transfer protocols to ensure data security; a data preprocessing unit that is operationally linked to the data acquisition unit and includes a data cleansing module for removing anomalies, duplicates and missing values, as well as a normalization engine for standardizing data formats, thereby ensuring a uniform data structure for processing; a predictive analytics unit configured to receive processed data from the data preprocessing unit, wherein the predictive analytics unit comprises an ensemble of machine learning models, including recurrent neural networks and long short-term memory networks, each model being optimized for different prediction horizons; an adaptive learning component embedded in the predictive analytics unit and configured to regularly retrain the machine learning models based on updated historical data and feedback on financial performance to ensure the accuracy of predictions through iterative improvement; a decision support unit that is operationally coupled with the predictive analytics unit and configured to provide real-time recommendations for cash allocation and liquidity adjustments based on forecasted cash flow patterns; a liquidity warning module configured to continuously monitor cash balances, compare them with pre-configured liquidity thresholds, and generate alerts when balances approach critical limits, thus enabling proactive management of cash flow risks; a visualization module configured to present insights into cash flow, trend forecasts, and liquidity recommendations in an interactive graphical format to facilitate user evaluation and strategic decision-making; and a network interface integrated into the system, configured to connect to external data sources for dynamic updates.

The invention relates to the technical field of image optimization, and discloses an image lossless fusion method and system based on depth feature extraction, and the method comprises the following steps: collecting multi-modal input data from different sensors, carrying out the preprocessing of each modal in the multi-modal input data, and carrying out the preprocessing of each modal in the multi-modal input data; the multi-modal input data comprises a visible light image, an infrared image and a depth map, a multi-layer pulse neural network is used for simulating pulse emission characteristics of biological neurons, time sequence feature extraction is performed on the multi-modal input data, and feature representations of different levels are captured. According to the method, deep feature extraction is carried out through the SNN algorithm to carry out image fusion, the problems of feature mutual exclusion and modal conflict when infrared and visible light images are processed through a traditional method can be effectively solved, a concrete image can be converted into an abstract feature, and then the abstract feature is converted into an end-to-end conversion process of the fused image. And the real-time and fidelity requirements in a dynamic scene are met.

The invention particularly relates to an underwater vehicle multi-energy hybrid energy storage system and a coordinated energy management strategy. The system comprises a lithium ion battery, a super capacitor and a fuel cell applying a proton exchange membrane which are respectively connected to a direct current bus. The method for coordinating the energy management strategy comprises the following steps: acquiring a current motion parameter and a current energy parameter of an underwater vehicle; inputting the current motion parameter, the current energy parameter and the expected navigational speed into a load prediction model based on a BP neural network, and estimating the expected total power at the next moment; and determining a corresponding action strategy according to the current operation mode of the underwater vehicle, and performing power distribution on the expected total power based on the action strategy. According to the scheme, the hybrid energy system can perform smooth working mode switching and optimal power distribution under different working conditions.

The invention provides a medical service item compliance dynamic monitoring method and system based on artificial intelligence, and relates to the technical field of medical service management.The method comprises the steps that a data incidence relation matrix is established through the unbalanced state entropy theory, and risk features are extracted through a chaotic neural network; a risk prediction spectrum is generated in combination with deep generative adversarial diagnosis and treatment modeling and heterogeneous medical network neural evolution analysis, and a causal reasoning decision tree is adopted to construct a time sequence evolution model to position abnormal risk points. The medical service project compliance real-time monitoring and risk early warning can be realized, the supervision efficiency is improved, and the medical risk is reduced.

The invention discloses a miniature sawtooth generation method based on an artificial intelligence optimization algorithm, and the method comprises the following steps: S1, initializing a generation parameter, and mapping the generation parameter to a high-dimensional complex field; s2, selecting and configuring an algorithm, and optimizing parameters by constructing a target function in a high-dimensional complex field; s3, constructing a neural morphology calculation system, processing the optimized parameters through a neural network and a complex modulation activation function, and generating a preliminary miniature sawtooth waveform; s4, error evaluation is carried out, an error signal is converted from a time domain to a frequency domain for weighted analysis and optimization, and key frequency band parameters are identified and adjusted; s5, adaptively adjusting the issuing frequency and synaptic weight of the neural network, generating parameters through continuous optimization, and circularly adjusting the waveform; and S6, outputting the optimized micro sawtooth waveform. According to the invention, high-precision and high-stability micro sawtooth waveform generation is realized, and the method is suitable for signal processing and application in a complex environment.

An intelligent, predictive customer loyalty system for social media marketing, consisting of: a data acquisition unit configured to collect user interaction data, demographic attributes, and sentiment analysis metrics from multiple social media platforms via platform-specific application programming interfaces (APIs), web scraping techniques, and direct data ingestion pipelines; a real-time engagement prediction engine comprising a machine learning model trained on historical engagement patterns, sentiment fluctuations, and content performance metrics, with the model dynamically predicting user responses, determining the optimal engagement time, and prioritizing interaction opportunities based on a probabilistic ranking technique; a real-time engagement prediction processing unit configured to generate and optimize social media posts, ads, and interactive media using a neural network-based natural language generation (NLG) model, where the model processes contextual user behavior and trending topics to dynamically adapt marketing content; an AI-powered automation unit for interactive engagement, comprising a deep learning-based chatbot engine integrated into a dialogue management system, with the chatbot engine analyzing the user's intentions, sentiment development, and engagement history to enable predictive, context-aware conversations with users; a cloud-connected processing unit that includes an AI inference accelerator configured to perform deep learning calculations for real-time engagement prediction units, personalized content recommendations, and chatbot response generation, with the processing unit integrated into a distributed computing framework to optimize inference latency.

System for determining the amount of sludge added to the boiler based on a coupled neural BP network, characterized in that it comprises the following: a creation module for the networked model of the combustion chamber, which serves to obtain a detailed structure of the denitrification system and the desulfurization system in the boiler, and to create a networked model of the combustion chamber based on the detailed structure of the denitrification system and the desulfurization system; a measurement module for the temperature field of the combustion chamber, which serves to create a relevant model for different coal qualities in the combustion chamber according to the networked model of the combustion chamber and in combination with the real-time data of the denitrification system and the desulfurization system, and to introduce CFD simulation values ​​into the relevant model and to simulate in order to obtain a temperature field of the combustion chamber; a denitrification efficiency acquisition module that serves to combine the detailed structure of the denitrification system with the temperature field of the combustion chamber and to obtain the denitrification efficiency and the NOx emission index within the framework of the 3D model; a measurement module for desulfurization efficiency, which serves to combine the detailed structure of the desulfurization system with the temperature field of the combustion chamber and to obtain the desulfurization efficiency and the SO2 emission index within the framework of the 3D model; a NOx dosing quantity module that serves to obtain the criterion for the minimum NOx emission value during boiler denitrification and the critical criterion for the NOx emission value with increasing sludge dosing quantity based on the denitrification efficiency and the NOx emission index within the framework of the 3D model, and to compare the different temperatures of the boiler chamber with the criterion of the minimum NOx emission value and the critical criterion of the NOx emission value with increasing sludge dosing quantity and to obtain the maximum regulated sludge dosing quantity corresponding to the NOx emission value; an SO2 dosing quantity module, which serves to obtain the criterion for the minimum SO2 emission value during the denitrification of the boiler and the critical criterion for the SO2 emission value with increasing sludge dosing quantity based on the desulfurization efficiency and the SO2 emission index within the framework of the 3D model, and to compare the different temperatures of the boiler chamber with the criterion of the minimum SO2 emission value and the critical criterion of the SO2 emission value with increasing sludge dosing quantity, and to obtain the maximum regulated sludge dosing quantity corresponding to the SO2 emission value; a model processing module that serves to train the maximum regulated sludge dosing quantity corresponding to the NOx emission value and the maximum regulated sludge dosing quantity corresponding to the SO2 emission value through a neural BP network and to obtain the maximum sludge dosing quantity of the boiler.

The disclosure includes a method (100) for classifying at least one detectable object (10) with at least one ultrasonic sensor (50). In this process, (120) object class information (12) for the at least one object (10) is determined using a neural network (9), wherein the neural network (9) generates an output (15) based on at least one first input (11), which indicates a probability value for at least one defined object class for the one object (10).

The invention relates to the technical field of unmanned vehicle management, in particular to an unmanned vehicle intelligent monitoring management system, which accurately detects electronic activities in a battery pack through a quantum tunneling effect electric quantity sensor, generates quantum signals, obtains multi-dimensional electric quantity data through processing, and provides a basis for energy management. Tensor analysis and a convolutional neural network are used to carry out deep mining on the multi-dimensional signal data, an energy flow state is monitored in real time, and an abnormal energy flow mode is identified; in combination with an energy flow analysis result, a charging scheduling scheme is generated and optimized by using chaotic mapping and a multi-objective optimization algorithm, and the charging efficiency, the vehicle waiting time and the power grid load are balanced; a charging adaptation model based on a biological neural network is constructed, intelligent charging adaptation decision making and charging equipment operation control are realized through reinforcement learning algorithm training, and the charging flexibility and efficiency are improved.

The invention discloses a novel power system primary frequency modulation optimization method based on deep learning, and relates to the technical field of power optimization. The method comprises the following steps: collecting state data of a hydroelectric generating set running in a power grid, constructing a three-dimensional feature tensor, and extracting spatio-temporal features through a TCN-GRU hybrid network; constructing a deep reinforcement learning decision layer, and defining an action space and a reward function; constructing a PID neural network with a time-varying forgetting factor, and outputting a dynamic weight; establishing a dual-time scale updating mechanism; a safety verification module is arranged, and when the system frequency deviation value is larger than a threshold value, the traditional PID mode is switched. Characteristics are extracted from state data of operation of a hydroelectric generating set in a power grid, the time sequence dependency relation of the characteristics is processed through deep learning, the frequency change trend of the power grid is captured, the optimal control strategy is explored by applying the DRL technology, PID controller parameters are adjusted in a dynamic environment, and the optimal control strategy is obtained. And the primary frequency modulation response speed and the control precision of the hydroelectric generating set during power grid frequency fluctuation are improved through self-adaptive control.

The invention discloses a closed-loop training device and a closed-loop training method for an in-vitro neural chip to control obstacle avoidance of a trolley, and belongs to the field of neurobiology and in-vitro brain-computer interfaces, and the closed-loop training device shapes a biological neural network structure and regulates and controls a discharge mode in a closed-loop training mode, so that the device makes a correct decision in different obstacle situations to control obstacle avoidance of the trolley. According to the device, by means of sensory electrical stimulation coding, the neuron culture can'sense 'the obstacle condition faced by the trolley in the environment, and then motion control over the external trolley is achieved by collecting and decoding the discharge signals of the neuron. According to the method, training electrical stimulation is adopted to provide feedback for a neural network, a system applies reward or punishment electrical stimulation to the neural network according to whether a trolley successfully avoids an obstacle or not, a synaptic structure and a discharge mode of the neural network are adaptively shaped and optimized, and the safety of the trolley is improved. And a network structure capable of autonomously controlling the trolley to avoid obstacles in different obstacle environments is gradually formed.

A system and method for real-time time series forecasting using a compound large codeword model with integrated supervisory neurons. The system processes diverse inputs through adaptive codebook generation and codeword allocation. A projection network fuses different data types for a latent transformer-based machine learning core. A hierarchical supervisory network, comprising low-level, mid-level, and high-level nodes, monitors local neural network regions, performing real-time statistical analysis and implementing structural modifications. The system efficiently handles multi-modal data, capturing complex relationships between input types. An adaptive codebook generation method, coupled with the supervisory architecture, ensures responsiveness to evolving data patterns and task requirements. This approach provides accurate and timely forecasts by leveraging diverse data types in a sophisticated, integrated manner, while continuously adapting its structure during operation to maintain optimal performance.

The invention discloses a method and device for automatic roasting and process intelligent control of a belt type roasting machine, and relates to the technical field of pellet roasting of the belt type roasting machine in the metallurgical iron and steel industry. The method comprises the four steps of historical data slice acquisition and modeling, target process parameter determination, feedback verification and online correction and rolling prediction control, a model historical library is established through service slices and time slices, modeling is performed based on a three-layer feed-forward neural network, flexible weighted deviation calculation combined with dynamic weight and rolling optimization of a softening coefficient are combined, and the rolling prediction control is performed. And accurate regulation and control of equipment parameters are realized. The automatic control system is mainly used for automatically controlling the pellet production process of the belt type roasting machine, the product quality stability is improved, the energy consumption is reduced, the labor intensity of workers is relieved, and support is provided for unmanned and intelligent production.

Disclosed is a method for determining a vehicle clutch temperature of a vehicle clutch by a neuron network. The method includes determining at least one input value representing a power supplied to the vehicle clutch, such that the at least one input value is determined on the basis of processing consecutive values of power supplied to the vehicle clutch. At least one input value and at least one value of an operating parameter are input as input data into the neuron network. A clutch temperature is determined by the neuron network on the basis of the input data and of a relationship learned by the neuron network between a time variation of the input data and the clutch temperature. A method is also disclosed for generating a training data set for a neuron network, and a control unit for determining a vehicle clutch temperature using a neuron network.

The present invention relates to a method for the computer-aided generation of a data-driven model for the computer-aided processing of digital SAR data from a repeat-pass SAR system configured to determine surface deformations. The digital SAR data consist of a plurality of interferograms (Γ), each of which is obtained from a first SLC acquisition (s1) of the SAR system at a first time point and a second SLC acquisition (s2) of the SAR system at a second time point later than the first, and a Doppler center frequency (fdc), an interferometric phase (φ̂), and a coherence (P), all of which are obtained from a respective pair of first and second SLC acquisitions (s1, s2). The data-driven model is configured to generate a zero-Doppler phase (φzd), which represents the phase contribution associated with the deformation causing a zero-Doppler LOS change.to determine an along-track phase (φat), representing the phase contribution associated with deformation in an along-track direction, and an estimate of an associated along-track surface motion (uat). The generation of the data-driven model comprises the following steps: - Generating a synthetic training dataset (TDS) comprising a variety of interferograms (Γ), each formed by a pair of a first and a second SLC capture (s1, s2), by inducing a model-based and external 3D deformation into a given first SLC capture (s1) to obtain the associated second SLC capture (s2), which relates to a time point after a deformation event, and by applying a reverse approach,which is specified by a given coherence (ρ) and a noise-free interferometric phase (φtheo) after the deformation event; - Training a neural network (NN) with the training dataset (TDS), wherein the neural network (NN) comprises an input for inputting the plurality of interferograms (Γ) and an output for outputting the along-track phase (Φat), the zero-Doppler phase (φzd) and the associated along-track surface motion (uat) for the plurality of interferograms (Γ), wherein the trained neural network (NN) represents the generated data-driven model.

The invention discloses a wireless communication network intelligent optimization method and system based on neuron collaboration, and relates to the technical field of wireless communication. The method comprises the following steps: mapping a communication node into a bottom layer sensing neuron and constructing a neural state variable set; constructing a node neural situation function based on the variables; when the local threshold value is exceeded, excitation pulses are generated and uploaded to middle-layer convergence neurons; the middle layer carries out pulse space aggregation and extreme value search, and outputs a selection strategy and an adjustment strategy; and reporting to a top layer to execute whole network neural synchronization index analysis, and optimizing the wireless communication network. The technical problems of low spectrum resource utilization rate and unstable network performance caused by the fact that a traditional wireless communication network cannot realize high-efficiency spectrum allocation and dynamic topology reconstruction under user mobility change are solved, and the purposes of realizing local quick response and global collaborative optimization by constructing a layered neural network and improving the network performance are achieved. And the spectrum resource utilization rate is improved, and the network dynamic adaptive capability is enhanced, so that the user service quality is guaranteed.

The invention discloses a multi-parameter micro-nano sensing chip for detecting a neural network. The multi-parameter micro-nano sensing chip comprises an ECIS interdigital electrode, an MEA multi-channel planar electrode, an electrochemical SWNT film array and a microfluidic structure. Three-dimensional heterogeneous integration is achieved through vertical stacking and partition cooperation, an ECIS interdigital electrode and an MEA chip share a plane base, an electrode layer is isolated through a nanoscale insulating layer, electrical signal crosstalk is avoided, meanwhile, the space of the chip is utilized to the maximum extent, meanwhile, a micro-fluidic cavity is divided into three corresponding detection areas, and the detection accuracy of the micro-fluidic cavity is improved. Therefore, signal detection of the in-vitro neural network in different states can be completed. According to the invention, multi-mode signal detection can be carried out on the neural network cultured by the same iPSCs neuron, and the ECIS interdigital electrode, the MEA multi-channel planar electrode and the electrochemical SWNT thin film array are combined together by using the micro-fluidic chip, so that the problems of single function, data splitting, insufficient sensitivity, difficulty in long-term stability and the like of the traditional neural chip are solved.

The embodiment of the invention discloses a business scene digital transformation method and device, equipment and a storage medium, and relates to the technical field of business scene digital transformation information. According to the scheme, digital plans of business scenes with different granularities are aggregated, features and resources of the business scenes are fused, multi-modal data are associated, key features are extracted, and the semantic comprehension ability of the digital plans and a large language model is intelligently matched; the neural network of the service scene is calculated according to the service scene attribution, the associated data and the positioning information, and the disposal scheme corresponding to the service scene is generated, so that subjective deviation of a traditional manual compiling solution is avoided, sudden working conditions and various service scenes are flexibly adapted, and the adaptability of complex scenes is enhanced.

The invention provides a method for indirectly measuring the carbon content of molten steel based on converter flue gas big data analysis, and belongs to the technical field of ferrous metallurgy. Comprising the following steps: starting measurement to obtain initial molten iron information; converter flue gas components are obtained; according to the change of the flue gas components, performing flue gas component theoretical correction on the obtained flue gas components; performing smoke component experience correction on the theoretically corrected smoke components according to offline neural network learning; calculating the oxidation amount of each element after the experience correction of the flue gas components; calculating the carbon content in the molten pool; correcting the calculated carbon content in the molten pool according to experience; judging whether a smelting end point is reached; if so, outputting an end point carbon content result; and if not, returning. The converter molten steel end point carbon is indirectly measured through converter flue gas big data analysis, manual carbon measurement of the converter can be replaced, the labor intensity is reduced, the cost is reduced, the labor productivity is improved, and intelligent steelmaking is promoted.

The invention discloses a parameter adaptive optical frequency comb repetition frequency phase locking system and method, and belongs to the field of optical frequency combs. The system comprises an optical fiber optical frequency comb, a repetition frequency detection module, a phase discriminator module, a parameter adaptive phase lock control module, a PZT frequency actuator and a frequency reference source module, the repetition frequency detection module is used for detecting a repetition frequency signal of the optical fiber optical frequency comb, and the frequency reference source module generates a target repetition frequency signal; the two paths of signals generate repetition frequency phase errors through the phase discriminator module, the repetition frequency phase errors enter the parameter self-adaptive phase lock control module, the parameter self-adaptive phase lock control module generates phase lock signals through a parameter self-adaptive phase lock controller, and control over repetition frequency of the optical fiber optical frequency comb is achieved through a PZT frequency actuator. The phase-locked controller is based on a single-neuron network PID controller, and can optimize PID control parameters in real time according to the current phase-locked condition so as to improve the phase-locked precision and the autonomous locking capability of the repetition frequency of the optical frequency comb in a complex dynamic environment.

The invention is suitable for the technical field of energy consumption control, and provides an energy consumption prediction and energy-saving control method adopting a deep neural network. According to the method, virtual training data is generated by constructing a machine room digital twinborn model including a three-dimensional geometric model and a CFD simulation model, and the electric energy use efficiency and the overheating condition are predicted in combination with a deep neural network optimized by an attention mechanism. And the operation strategy system generates an optimal control suggestion by predicting the energy-saving effect and the safety risk after the air conditioner is turned off or the temperature is increased. The visual interface integrates real-time parameters, historical data and three-dimensional airflow organization dynamic display, and assists an administrator in decision making. The method breaks through the limitation of real data, improves the prediction precision, reduces the PUE on the premise of ensuring that the operation number of air conditioners is greater than or equal to 3, and achieves the efficient and energy-saving management of a machine room. Experiments show that the PUE can be reduced by about 4.3%, and the cabinet overheating probability can be reduced by 60%.

The invention relates to the technical field of power electronics, and discloses a dynamic regulation and control method and system for an active clamping three-level converter adapting to working condition changes, and the method comprises the steps: calculating the loss sample data of each power device in different modulation modes through obtaining the historical data of characteristic parameters under different working conditions; calculating the junction temperature of the power device based on the loss in the hybrid modulation mode; with the minimum junction temperature distribution imbalance degree as a target, calculating a hybrid modulation coefficient, and constructing a neural network model; after operation parameters of a power device are collected in real time, the optimal modulation coefficient is obtained based on the model, the modulation mode of the converter is dynamically adjusted, and loss distribution is optimized. According to the method, the loss distribution characteristic and efficiency of the converter under all working conditions can be improved, and the loss balance performance is optimized.

An active electro-physiological system and method are provided. The present technology is directed to use a multi electrode array (MEA) chip to emit a temporal-spatial patterned stimulation signal data to a neuronal tissue (e.g., a brain organoid) and receive response signal data from the neuronal tissue. The system actively determines a representational function of a neuronal tissue based on correlation between temporal-spatial patterned stimulations and responses. The system performs functional phenotyping according to chemical and / or biological reagents as an external input in contact with the neuronal tissue. The active electro- physiological system diagnoses a disease by using cells and tissues of a clinical patient as the neuronal tissue placed in the MEA chip and / or as an external input to the neuronal tissue. The active electro-physiological system further performs biocomputing by training the neuronal tissue to classify input data and / or predict output data according to a representational function of the neuronal tissue.

A system (100) for context-sensitive language translation, comprising: an input interface module (102) configured to allow the user to input text for language translation, wherein the input interface module (102) includes an interface that allows the user to input text and select the language for translation; a context processing module (104) connected to the input interface module (102) and configured to enable context-dependent extraction of information from the provided input text, wherein the context processing module (104) comprises the following: an input processing module (106) configured to decompose the input text into tokens and identify context-sensitive segments; and a context extraction module (108) connected to the input processing module (106), wherein the context extraction unit (108) comprises an extraction processor configured to generate context vectors from source input text; a translation processing unit (110) comprising a machine learning-based processor configured to translate the provided input text, wherein the translation processing unit comprises: an attention mechanism (112) configured to select words based on their relevance to the context; and a translation generation module (114) configured to generate translations using a neural network that incorporates the attention-weighted context vectors; a post-processing unit (116) connected to the translation processing unit (110) and configured to refine the translated output to ensure fluency and coherence, refining the translation grammar, ensuring semantic consistency, validating contextual appropriateness, and generating a fluency score; and an output module (118) connected to the post-processing unit (116) and comprising a display with an output interface on which the input text and the translated output text are displayed to the user.

The invention discloses an intelligent detection navigation system and method for an inland ship. The method comprises the following steps: step 1, acquiring multi-modal sensing data in the navigation process of the inland ship; 2, constructing an initial cognitive map; step 3, carrying out dynamic modeling on each semantic state node by adopting an improved liquid neural network, embedding symbolic logic constraints in state transition, and generating a navigation control bias steering quantity; step 4, calculating a tension value of a causal directed edge and executing a structure evolution operation to generate a self-evolution cognitive map; 5, constructing a candidate path set; step 6, calculating an atlas winding degree index of each candidate path; 7, constructing a path decision vector, executing a path backstepping subduction operation, and screening to obtain a target optimal path; and 8, converting the target optimal path into a navigation control instruction, and outputting the navigation control instruction to a ship control system. According to the invention, cognitive map modeling and neural network reasoning are fused, and intelligent navigation control of inland ships is realized.

The invention relates to a method (100, 200), a device, a neural network (DNN), in particular a Generative Pre-trained Transformer (GPT), and a computer program for operating (106) a vehicle (EgoF). The method (100, 200) comprises: - Determine (101) the action data that are characteristic of several automated actions (AA1, AA2, AA3) of the vehicle (EgoF); and: - Operating (203) a neural network (DNN, GPT) depending on the data based on the action data, generating an evaluation information (104) that takes into account a legal and / or moral evaluation of a result of one or more automated actions (AA1, AA2, AA3), in particular characterizing; and: - Effect (106, 107) one or more actions depending on the assessment information.

Method for training an artificial neural network (ANN), wherein the method exhibits: Storing weight values ​​in an array of resistive processing units (12), RPU, wherein the array of RPU units (12) represents a weight matrix (10) W of the ANN with m rows and n columns, by storing the weight values ​​of the weight matrix (10) W as resistive values ​​of the RPU units in the array; Define the weighting matrix (10) W such that it has an output dimension that is smaller than the input dimension, so that the weighting matrix (10) W has a rectangular shape; and Transforming (1010) the weighting matrix (10) W from a rectangular shape to a more square shape by repeating or concatenating the rectangular shape of the weighting matrix (10) W to increase the signal strength of a reverse pass signal by: Copying (1020) an input of repeated weighting elements during a forward cycle pass; Summing (1030) output calculations from the repeated weighting elements, resulting in one output per line; and Update (1040) each of the repeated weight elements according to a backward propagated error or alternatively update (1040) only one of the repeated weight elements by setting all forward values ​​except one to zero during an update pass.

The present invention relates to a method for determining the concentration of a chemical compound in a fluid, by means of an optical measurement system for measuring a light intensity spectrum, such that: (i) by means of a training set comprising a plurality of measured training light intensity spectra for a plurality of training fluids of predetermined concentration, a model is built to determine the concentration of at least one chemical compound in a fluid by training a one-dimensional convolutional neural network on the training set, the network being such that the number of convolutional filters per layer decreases with the position of the layer in the network; (ii) by means of the optical measurement system, at least one light intensity spectrum of the fluid is measured; (iii) the constructed model is applied to the measured light intensity spectrum, and the concentration of the chemical compound in the fluid is determined.Figure 2 to be published.