322 results about "Computation graph" patented technology

The present invention extends to methods, systems, and computer program products for recognizing input gestures. A neural network is trained using example inputs and backpropagation to recognize specified input patterns. Input gesture data is representative of movements in contact on a multi-touch input display surface relative to one or more axes over time. Example inputs used for training the neural network to recognize a specified input pattern can be created from sampling input gesture data for example input gestures known to represent the specified input pattern. Trained neural networks can subsequently be used to recognize input gestures that are similar to known input gestures as the specified input pattern corresponding to the known input gestures.

A system for improving a distributable model with biases contained in distributed data is provided, comprising a network-connected distributable model configured to serve instances of a plurality of distributable models; and a directed computation graph module configured to receive at least an instance of at least one of the distributable models from the network-connected computing system, create a cleansed dataset from data stored in the memory based at least in part by biases contained within the data stored in memory, train the instance of the distributable model with the cleansed dataset, and generate an update report based at least in part by updates to the instance of the distributable model.

A computer-implemented method of neural architecture search is disclosed. Embodiments may provide a neural network (100) configured to perform a selected task. A computational graph (200) is obtainedwhich includes a plurality of nodes, edges and weightings associated with the nodes / edges. The computational graph (200) includes a plurality of candidate models in the form of subgraphs (301,302) ofthe computational graph (200). Selected subgraphs are trained sequentially, with the weightings corresponding to each said subgraph being updated in response to training. For each weighting in a subgraph which is shared with another subgraph, updates to the weightings (in response to training that subgraph) are controlled based on an indication of how important to the another subgraph a node / edgeassociated with that weighting is.

The invention discloses a high-resolution construction land graph spot identification method based on PanTex and a linear characteristic. The method comprises the following specific steps: step one, performing registering on high-resolution remote sensing images and corresponding land utilization graph spots; step two, performing superposing mask on the land utilization graph spots and the high-resolution remote sensing images to obtain independent graph spot images corresponding to graph spot polygons; step three, calculating the PanTex characteristic images of the high-resolution remote sensing images after processing, and counting the sum of PanTex indexes in each graph spot; step four, extracting straight lines in graph spot images for the independent graph spot images after the processing; step five, calculating the linear characteristics of the graph spots; and step six, classifying the graph spots by use of two types of SVM classifiers, and extracting construction land graph spots. According to the invention, the problem of failure of the PanTex indexes in large workshops and large roofs in the high-resolution remote sensing images is solved, the algorithm is simple and highly-efficient, the result is in the form of the graph spots, and a GIS database can be conveniently updated.

The invention provides a resource scheduling method. The method comprises the steps of obtaining a job program of a deep learning job, and converting the job program to obtain an intermediate representation of a calculation graph; segmenting the intermediate representation of the calculation graph to obtain a sub-graph set; packaging the sub-graph set to obtain workload mirror images correspondingto various accelerators; and determining a target accelerator from the accelerator cluster according to preset accelerator capability information, the service level condition submitted by the user and the information of the resource pool, and sending a corresponding workload mirror image to the target accelerator. According to the method, unified abstraction is carried out on operation programs of different frameworks by utilizing intermediate representation of the calculation graph; and based on the intermediate representation of the calculation graph, multiple workload mirror images, comprehensive accelerator capability information, service level conditions and information of a resource pool are obtained, a target accelerator is determined, corresponding workload mirror images are allocated to the target accelerator, accelerator resources are reasonably utilized, and the use efficiency is improved. The invention provides a resource scheduling device and system with the above beneficial effects.

The embodiment of the invention provides a configurable heterogeneous artificial intelligence processor. The configurable heterogeneous artificial intelligence processor comprises at least two calculation units of different structure types, task queues, storage units and controllers, wherein the task queues, the storage units and the controllers correspond to the calculation units. Each controllerdecomposes a calculation graph of a to-be-processed neural network into a plurality of calculation subtasks and distributes the calculation subtasks to corresponding task queues of all calculation units, and the dependency relationship between all the calculation subtasks is set. According to the set dependency relationship, synchronization among the calculation sub-tasks and controlling access of data related to the calculation sub-tasks is achieved to the storage units and an off-chip memory. According to the technical scheme of the embodiment of the invention, an on-chip heterogeneous formis adopted, and the single controller uniformly schedules and manages the calculation units of each architecture to process different application tasks, so that an artificial intelligence processor can flexibly adapt to different application scenes, the expandability is improved, and the efficiency of processing different tasks is improved.

A system for cyber threat hunting employing an advanced cyber decision platform comprising a time series data store, a directed computational graph module, an automated planning service module, and observation and state estimation module, wherein the state of a network is monitored and used to predict network resources that may be vulnerable to a future cyber threat and to produce a cyber-physical graph representing the vulnerable network resources, a human operator is provided with the cyber-physical graph to analyze the data contained therein to initiate an investigation of network resources, and the results of the threat investigation and their effects are analyzed to produce security recommendations.

The invention relates to a neural network compiler configuration method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a target neural network model and an initial calculation graph corresponding to the target neural network model, wherein the initial calculation graph comprises a plurality of operators; dividing the plurality of operators to obtain a plurality of operator sets; determining a plurality of operator type combinations respectively corresponding to each operator set, and obtaining operator operation time corresponding to each operator type combination to obtain a plurality of operator operation times respectively corresponding to each operator set; taking the operator type combination with the shortest operator operation time in the plurality of operator type combinations as a target operator type combination corresponding to each operator set; and generating a target calculation graph of the target neural network model according to the target operator type combination corresponding to each operator set, and generating compiler configuration information for the target neural network model according to the target calculation graph. By adopting the method, the configuration optimization effect of the neural network model compiler can be improved.