17 results about "Weight space" patented technology

The invention discloses a scene adaptive recommendation method in smart home based on multi-modal data, and relates to the technical field of multi-modal data processing. According to the method, firstly, scene classification is carried out based on multi-mode sensing data for a smart home scene, a current life scene of a user is accurately recognized, meanwhile, an equipment linkage state is analyzed in combination with a classification result, whether equipment operation is matched with a scene requirement or not is judged, and then a scene mode is recognized according to a linkage analysis result to generate preliminary recommendation; whether the dynamic weight space relation model of the user is updated or not is judged, finally, the recommendation effect is quantified according to the weight adjustment judgment result, the recommendation strategy is optimized through two-way feedback, the adaptability to the user requirement is improved, and the user experience is improved. A full-process closed loop of multi-modal data classification, equipment linkage analysis, mode recognition recommendation, weight dynamic updating and effect quantitative feedback in a smart home scene is realized, and the problem that the intelligent degree of scene recommendation in smart home is not high in a user behavior dynamic change scene is effectively solved.

Techniques of nonlinear quantization of an artificial neural network model having first weights. For example, a predetermined number of unique, second weights having a nonlinear distribution in a weight space of the first weights can be identified to generate a quantized model based on replacing, in the artificial neural network model, the first weights with closest ones from the second weights. A linear mapping between the second weights and values of conductance of memristors of an accelerator configured to perform operations of multiplication and accumulation can be used to determine the same predetermined number of programming voltages. Conductance of the memristors can be programmed using the programming voltages in preparation of the accelerator to perform an operation of multiplication and accumulation in the quantized model. The nonlinear distribution and the linear mapping can be adjusted to increase or optimize the accuracy of the quantized model.

The application provides a ship radiation noise recognition method and system based on a double low-rank adjustment network. The ship radiation noise recognition method based on the double low-rank adjustment network comprises the following steps: performing Mel filter bank transformation on a ship radiation noise signal to obtain a Mel time-frequency spectrum diagram of the noise signal; inputting the Mel time-frequency spectrum diagram into a trained noise recognition model to output a target category of the ship radiation noise; and generating the noise recognition model based on a pre-trained benchmark model containing a plurality of residual blocks, performing low-rank adaptation on a weight space through a weight low-rank adaptive module, and performing low-rank adaptation on a feature space through a feature low-rank adaptive module. The application has the advantages that the classification performance can be significantly improved on ShipsEar and DeepShip data sets by increasing only about 0.55% parameters, and the application has low overhead and high effectiveness.

The application relates to a control method for dynamic walking of a biped robot and the biped robot, and belongs to the technical field of robot control, which comprises the following steps: 1, a self-recurrent cerebellar model neural network is used to establish a dynamic model of the biped robot with a disturbance term, and dynamic robust walking of the biped robot is converted into a problem of realizing stability of a multi-input multi-output nonlinear system with a bounded uncertain term; 2, an adaptive self-recurrent cerebellar model neural network error observer is designed to estimate an error upper limit; 3, an adaptive law of network weight is designed to realize real-time updating of the network weight space and to adjust parameters of each learning factor; and 4, a boundary value estimation algorithm is used to compensate for an estimation error and feedback to a robot walking system, so that the biped robot can realize asymptotic stable walking. The application enables the control system to adapt to time-varying characteristics of the biped walking system on line, and has continuous learning and adaptive capacity for unknown dynamics.

The invention provides a target detection method, device and equipment. According to the method provided by the invention, feature extraction is carried out on a prompt text of a to-be-detected image and a prompt text of a to-be-detected target, and feature maps and text features of multiple scales are obtained; generating a self-attention feature of the text feature through a self-attention mechanism; for the feature map of each scale, mapping the feature map of the scale to a weight space of a convolution kernel, and generating a weight of a dynamic convolution kernel corresponding to the scale; performing convolution processing on the feature map of the scale by using the dynamic convolution kernel corresponding to the scale to obtain a dynamic convolution feature under the scale; and performing target detection by using the plurality of dynamic convolution features under the plurality of scales and the self-attention features to obtain a detection result. According to the target detection method, device and equipment provided by the invention, the calculation amount of target detection can be reduced, so that the applicability of a target detection algorithm is improved, and the target detection algorithm is applied to edge low-calculation-power equipment.

The application relates to the technical field of electric power Internet of Things and data processing, and particularly discloses a multi-dimensional data center construction system of adjustable resources based on a measurement center, which comprises the following steps: collecting electrical measurement data of adjustable resources in high precision by an edge collection terminal; generating candidate edges by fusing the correlation and power transfer entropy of the measurement data; obtaining a weighted space-time correlation graph by combining a graph attention network; performing Laplace interpolation on missing data by using the graph and outputting a confidence degree; identifying a low-confidence degree area according to the spatial distribution of the confidence degree, and dynamically optimizing a collection frequency and a cleaning strategy; generating a virtual resource cluster with electrical close coupling by adopting hierarchical clustering based on the edge weight of the graph; and constructing an adjustable feasible region and optimizing the distribution of regulation and control instructions by taking the edge weight of the graph as a constraint in the cluster, so that the problems of topology dynamic tracking, data quality enhancement, resource aggregation rationality and regulation and control collaboration are solved.

The invention relates to the technical field of multi-source data fusion, in particular to a project management method based on multi-source monitoring data fusion, and the method improves the data fusion precision, efficiency and query speed, quantifies the reliability difference of monitoring means based on a CRITIC method, introduces an actual availability screening and time index decay model, and improves the data fusion accuracy, efficiency and query speed. Weight space-time dual dynamic optimization is realized, decision deviation is overcome, an objective basis is provided for overlapped pattern spot judgment, and fusion accuracy and credibility are improved. Weight updating is controlled through the QZGXSJ field, total data recalculation is avoided, resource consumption is reduced, the design is compatible with mainstream GIS database standards, seamless connection with an existing system is achieved, and popularization is easy. Meanwhile, global data are stored according to quarterly slices by using a time axis engine, and a project engine is evolved according to a code storage version, so that joint storage is realized, the problem of fragmented management is solved, and the query speed is increased.

The application discloses an Internet of Things data mining method and device based on machine learning and a medium, relates to the technical field of data mining, and comprises the following steps: constructing a space-time joint attention model, calculating time step attention weights and inter-device space relationship attention weights, dynamically adjusting the fusion ratio of the time step attention weights and the inter-device space relationship attention weights by evaluating the time variation rate and the space density of the preprocessed data stream, generating weighted space-time feature representation, initializing a basic model through a lightweight online learning architecture, continuously learning the weighted space-time feature representation by using an incremental update algorithm, adjusting the basic model parameters, and generating a real-time optimized learning model; by means of the space-time joint attention model, the time step attention weights and the inter-device space relationship attention weights are fused, unified feature representation of time dependence and space correlation is realized, and the mining capacity of complex space-time correlation is improved.

The present application relates to the technical field of data analysis, in particular to a method for determining the regional boundary of territorial space planning based on multidimensional data, in which the geographical space, social economy and ecological sensitive attributes are uniformly mapped through the rule space unit grid to eliminate the scale misplacement and boundary jump problems of multi-source data; the attribute mutual exclusion conflict nodes are accurately positioned based on the gradient direction deviation detection, and the dynamic coupling network is constructed by using the non-conflict nodes to quantify the spatial transmission relationship; the dynamic connection weight is globally converged through the iterative compression operation to break through the limitation of the static model and inhibit the local conflict interference; finally, the closed boundary line is output under the premise of maintaining the spatial continuity by relying on the topological deformation driven by the weight spatial gradient, the problems of resource continuity fragmentation and attribute logic fracture are solved, and the multi-source coordinated adaptation of the planning boundary and the natural-economy-ecological gradient is realized.

The application discloses a wide-area multi-bus load prediction method based on a gated space-time graph neural network, determines weather features strongly related to bus loads through a fast maximum information coefficient, determines time-space coupling correlations among the bus loads through the fast maximum information coefficient, and completes construction of a similarity weight space-time graph through the determined weather features, extracts and mines spatial features of each node of the similarity weight space-time graph in a graph convolution mode, inputs a time sequence formed by a result of a space convolution layer into a gated recurrent unit layer, and realizes time domain feature mining through the gated recurrent unit, so that the problems that the influence of unstructured time-space coupling correlations among the multi-bus loads in a wide-area space on a prediction result is not fully considered and it is difficult to uniformly model multi-bus load prediction are solved, global multi-node feature enhancement is realized, and load prediction precision is effectively improved.

The invention discloses a GIS data rapid retrieval and visualization method fusing spatio-temporal characteristics, and relates to the technical field of spatio-temporal geographic information systems.The method comprises the steps that a dynamic grid is divided according to data distribution entropy, and a spatio-temporal association index is constructed through a 3D-Hilbert algorithm; weighting the spatial-temporal characteristics by adopting an entropy weight method, and converting a query range into coding intervals for screening and sorting; dynamically adapting a visualization strategy according to the result density; the rendering precision is adjusted according to the camera distance and the screen error, and strategies are updated and switched in real time; newly-added data is inserted or a cube is newly built according to needs, index balance is maintained, and reconstruction is carried out; a space-time association network is constructed, and association rules are mined by means of a community discovery algorithm. Through innovative index construction, feature fusion retrieval and dynamic visualization, the GIS data retrieval efficiency and accuracy are improved, the visualization effect is optimized, data updating and deep analysis are supported, interactivity is enhanced, and the data processing value is improved in multi-field application.

A memory module system for a high-dimensional weight space neural network configured to process machine learning data streams using Bayesian Inference and / or Classical Inference is set forth. The memory module can include embedded high speed random number generators (RNGs). The memory module is configured to compute, store and sample neural network weights by adapting operating precision to optimize the computing effort based on available weight space and application specifications.

An illustrative example embodiment of a counterweight monitoring device includes a boundary marker configured to be located at a selected location along a counterweight path along a boundary of a counterweight space. The detector is configured to detect a boundary marker change. The detector is configured to provide an output indicative of a deviation of any portion of the counterweight from the counterweight space based on the detected boundary mark change.

The present application belongs to the field of computer vision, and relates to a real-time three-dimensional human pose estimation method and device oriented to space-time topology modeling. The method comprises: using a contrast learning paradigm for unlabeled pre-training to extract human topology structure prior information; using a light-weight space-time human topology extraction network as an encoder, using human topology structure prior information, and using a double-flow structure to respectively and parallelly extract human kinematic topology relations in a time domain and human geometric topology relations in a space domain; performing accumulation operation on the human kinematic topology relations in the time domain and the human geometric topology relations in the space domain, and obtaining positions of human three-dimensional poses through linear regression and outputting corresponding three-dimensional skeleton coordinates. The present application can accurately capture human structure and kinematic topology features, and can realize efficient and accurate three-dimensional human pose estimation.

The invention relates to the technical field of intelligent driving, in particular to a heterogeneous multi-agent space-time distribution trajectory planning method based on congestion driving, and the method comprises the steps: obtaining a data set composed of physical parameters, motion performance limit data and task target data of heterogeneous agents, establishing a dynamic model of the heterogeneous intelligent agent based on the data set, constructing a Gaussian weighted space-time grid according to the track occupation set, extracting real-time environment characteristics of the intelligent agent by utilizing space-time congestion degree data output by the space-time congestion degree prediction model, and constructing a scene adaptive weight adjustment mechanism according to the environment characteristics; dynamically optimizing the target function of trajectory planning; on the basis of the optimized objective function, establishing a half-space constraint relation between the intelligent agents, and on the basis of the distributed asynchronous architecture, generating and executing the optimal trajectory, the boundary can be automatically adjusted when the environment is crowded and changes or the intelligent agents deviate from the predicted trajectory, and the trajectory quality and the system efficiency are remarkably improved.

This application provides a scenario-oriented method, system, device, medium, and program product for large-scale model reconstruction, relating to the field of large-scale model training. It obtains a business task description and isomorphic first and second models; extracts importance parameters from the layer-by-layer parameter differences between the two models; performs Bayesian optimization in the layer-by-layer fusion weight space, utilizes a probabilistic surrogate model to model the mapping relationship between capability and efficiency, outputs candidate weights with a Pareto front quality-oriented acquisition strategy, continuously improves the front strength and density, and outputs candidate fusion models and offline evaluation records after meeting the required number of evaluations; finally, it selects the optimal model for execution based on hard constraints on inference budget and a preference strategy. This application can maximize inference throughput, reduce service costs, and improve system response speed and user experience while ensuring task accuracy.