Resumen de: US2023013006A1

The invention relates to a system for monitoring and controlling a dynamic network such as an oil, gas, or water pipeline. The system includes a plurality of sensors for measuring aspects of a state of the network with each sensor being associated with a segment of the network and connected to a virtual sensor which accumulates and pre-processes measurements from the sensors for each segment of the network. The system further includes a network topology processor for storing the topology of the network and relating sensors and virtual sensors to segments of the network and neighbouring sensors and virtual sensors in accordance with the topology and a reinforcement learning artificial neural network (ANN) based nonlinear state estimation and predictive control model which uses measurements from the sensors and virtual sensors to model the state of the network and estimate sequential states of the network.

Resumen de: US2023018248A1

Embodiments of the present disclosure relate to a neural engine of a neural processor circuit having multiple multiply-add circuits and an accumulator circuit coupled to the multiply-add circuits. The multiply-add circuits perform multiply-add operations of a three dimensional convolution on a work unit of input data using a kernel to generate at least a portion of output data in a processing cycle. The accumulator circuit includes multiple batches of accumulators. Each batch of accumulators receives and stores, after the processing cycle, the portion of the output data for each output depth plane of multiple output depth planes. A corresponding batch of accumulators stores, after the processing cycle, the portion of the output data for a subset of the output channels and for each output depth plane.

Resumen de: US2023017304A1

A mechanism is described for facilitating smart distribution of resources for deep learning autonomous machines. A method of embodiments, as described herein, includes detecting one or more sets of data from one or more sources over one or more networks, and introducing a library to a neural network application to determine optimal point at which to apply frequency scaling without degrading performance of the neural network application at a computing device.

Resumen de: US2023014465A1

Systems and methods for efficiently identifying and extracting machine-actionable structured data from web documents are provided. The technology employs neural network architectures which process the raw HTML content of a set of seed websites to create transferable models regarding information of interest. These models can then be applied to the raw HTML of other websites to identify similar information of interest. Data can thus be extracted across multiple websites in a functional, structured form that allows it to be used further by a processing system.

Resumen de: AU2021289232A1

This disclosure relates to detecting visual findings in anatomical images. Methods comprise inputting anatomical images into a neural network to output a feature vector and computing an indication of visual findings being present in the images by a dense layer of the neural network that takes as input the feature vector and outputs an indication of whether each of the visual findings is present in the anatomical images. The neural network is trained on a training dataset including anatomical images, and labels associated with the anatomical images and each of the visual findings. The visual findings may be organised as a hierarchical ontology tree. The neural network may be trained by evaluating the performance of neural networks in detecting the visual findings and a negation pair class which comprises anatomical images where a first visual finding is identified in the absence of a second visual finding.

Resumen de: US2023012645A1

In various examples, a deep neural network (DNN) is trained for sensor blindness detection using a region and context-based approach. Using sensor data, the DNN may compute locations of blindness or compromised visibility regions as well as associated blindness classifications and/or blindness attributes associated therewith. In addition, the DNN may predict a usability of each instance of the sensor data for performing one or more operations—such as operations associated with semi-autonomous or autonomous driving. The combination of the outputs of the DNN may be used to filter out instances of the sensor data—or to filter out portions of instances of the sensor data determined to be compromised—that may lead to inaccurate or ineffective results for the one or more operations of the system.

Resumen de: US2023017503A1

The present disclosure provides an artificial intelligence (AI) system for sequence-to-sequence modeling with attention adapted for streaming applications. The AI system comprises at least one processor; and memory having instructions stored thereon that, when executed by the processor, cause the AI system to process each input frame in a sequence of input frames through layers of a deep neural network (DNN) to produce a sequence of outputs. At least some of the layers of the DNN include a dual self-attention module having a dual non-causal and causal architecture attending to non-causal frames and causal frames. Further, the AI system renders the sequence of outputs.

Resumen de: US2023016670A1

An information processing apparatus includes a memory and processor. The memory stores a first inference model using a neural network and a plurality of defense algorithms. The at least one processor performs acquisition of prescribed data, input of the prescribed data to the first inference model to perform inference processing, the first inference model being learned using learning data including respective data and respective result data obtained by solving prescribed problems using the respective data, detection of a possibility as to whether a prescribed attack has been made on the prescribed data, specification of, when the possibility of the prescribed attack is detected, a first defense algorithm capable of making a defense against the prescribed attack from among the plurality of defense algorithms on a basis of the prescribed data on which the prescribed attack has been made, and application of the first defense algorithm to the inference processing.

Resumen de: US2023012843A1

An autonomous driving system for a vehicle reduces the amount of computations for object extraction carried out by a DNN, using information a traveling environment or the like. An information processing apparatus including a processor, a memory, and an arithmetic unit that executes a computation using an inference model is provided. The information processing apparatus includes a DNN processing unit that receives external information, the DNN processing unit extracting an external object from the external information, using the inference model, and a processing content control unit that controls processing content of the DNN processing unit. The DNN processing unit includes an object extracting unit that executes the inference model in a deep neural network having a plurality of layers of neurons, and the processing content control unit includes an execution layer determining unit that determines the layers used by the object extracting unit.

Resumen de: WO2021248140A1

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating an ensemble of neural networks. In particular, the neural networks in the ensemble are trained using different hyperparameters from one another.

Resumen de: US2023009282A1

A method includes receiving a video signal that comprises a time series of images of a face of a human, wherein the images in the time series of images comprise a set of landmark points in the face, applying tracking processing to the video signal to reveal variations over time of at least one image parameter at the set of landmark points in the human face, generating a set of variation signals indicative of variations revealed at respective landmark points in the set of landmark points, applying processing to the set of variation signals, the processing comprising artificial neural network processing to produce a reconstructed PhotoPletysmoGraphy (PPG) signal, and estimating a heart rate variability of a variable heart rate of the human as a function of the reconstructed PPG signal.

Resumen de: US2023008443A1

A system processes images of documents, for example, identification documents. The system transforms an image of a document to generate an image that represent the document in a canonical form. For example, if the input image has a document that is tilted at an angle with respect to the sides of the image, the system modifies the orientation of the document to show the document having sides aligned with the sides of the image. The system stores user accounts that include user information including images. The system generates a graph of nodes that represent user accounts with edges determined based on similarity scores between user accounts. The system determines connected components of user accounts, such that each connected component represents user accounts that have a high likelihood of being duplicates.

Resumen de: WO2021178009A1

Described is a system for proving correctness properties of a neural network for providing estimates for point cloud data. The system receives as input a description of a neural network for generating estimates from a set of point cloud data. The description of the neural network is parsed to obtain a symbolic representation. Based on a combination of the symbolic representation and a set of analysis parameters, the system generates an analysis output indicating whether the neural network satisfies a correctness property in generating the estimates from the set of point cloud data. The analysis output is a mathematical proof artifact proving that the set of analysis parameters is satisfied, a list of one or more point clouds for which the set of analysis parameters is violated, or a report that progress could not be made by the analysis.

Resumen de: WO2021178872A1

Various embodiments of the technology described herein generally relate to systems and methods for trajectory optimization with machine learning techniques. More specifically, certain embodiments relate to using neural networks to quickly predict optimized robotic arm trajectories in a variety of scenarios. Systems and methods described herein use deep neural networks to quickly predict optimized robotic arm trajectories according to certain constraints. Optimization, in accordance with some embodiments of the present technology, may include optimizing trajectory geometry and dynamics while satisfying a number of constraints, including staying collision-free and minimizing the time it takes to complete the task.

Resumen de: WO2021178865A1

Various embodiments of the technology described herein generally relate to systems and methods for trajectory optimization with machine learning techniques. More specifically, certain embodiments relate to using neural networks to quickly predict optimized robotic arm trajectories in a variety of scenarios. Systems and methods described herein use deep neural networks to quickly predict optimized robotic arm trajectories according to certain constraints. Optimization, in accordance with some embodiments of the present technology, may include optimizing trajectory geometry and dynamics while satisfying a number of constraints, including staying collision-free, and minimizing the time it takes to complete the task.

Resumen de: WO2021178009A1

Described is a system for proving correctness properties of a neural network for providing estimates for point cloud data. The system receives as input a description of a neural network for generating estimates from a set of point cloud data. The description of the neural network is parsed to obtain a symbolic representation. Based on a combination of the symbolic representation and a set of analysis parameters, the system generates an analysis output indicating whether the neural network satisfies a correctness property in generating the estimates from the set of point cloud data. The analysis output is a mathematical proof artifact proving that the set of analysis parameters is satisfied, a list of one or more point clouds for which the set of analysis parameters is violated, or a report that progress could not be made by the analysis.

Resumen de: US2023004204A1

Systems, apparatuses, and methods for managing power consumption for a neural network implemented on multiple graphics processing units (GPUs) are disclosed. A computing system includes a plurality of GPUs implementing a neural network. In one implementation, the plurality of GPUs draw power from a common power supply. To prevent the power consumption of the system from exceeding a power limit for long durations, the GPUs coordinate the scheduling of tasks of the neural network. At least one or more first GPUs schedule their computation tasks so as not to overlap with the computation tasks of one or more second GPUs. In this way, the system spends less time consuming power in excess of a power limit, allowing the neural network to be implemented in a more power efficient manner.

Resumen de: US2023004796A1

Systems and methods are described for developing and using neural network models. An example method of training a neural network includes: oscillating a learning rate while performing a preliminary training of a neural network; determining, based on the preliminary training, a number of training epochs to perform for a subsequent training session, and training the neural network using the determined number of training epochs. The systems and methods can be used to build neural network models that efficiently and accurately handle heterogeneous data.

Resumen de: US2023004804A1

Systems, devices, and methods related to a deep learning accelerator and memory are described. An integrated circuit may be configured with: a central processing unit, a deep learning accelerator configured to execute instructions with matrix operands; random access memory configured to store first instructions of an artificial neural network executable by the deep learning accelerator and second instructions of an application executable by the central processing unit; one or connections among the random access memory, the deep learning accelerator and the central processing unit; and an input/output interface to an external peripheral bus. While the deep learning accelerator is executing the first instructions to convert sensor data according to the artificial neural network to inference results, the central processing unit may execute the application that uses inference results from the artificial neural network.

Resumen de: US2023004880A1

A method for optimizing production of sheet-metal parts, the production comprising cutting out and singularizing the sheet-metal parts and bending the sheet-metal parts, wherein the method includes: (A) training a neural network, which is executed on a Monte Carlo tree search framework, by means of supervised learning and self-play with reinforcement learning; (B) recording constraints for the sheet-metal parts, the constraints comprising geometric data of the sheet-metal parts; (C) creating an optimized production schedule by way of the neural network; and (D) outputting the production schedule.

Resumen de: WO2023276251A1

The present disclosure provides an artificial intelligence (AI) system for sequence-to-sequence modeling with attention adapted for streaming applications. The AI system comprises at least one processor; and memory having instructions stored thereon that, when executed by the processor, cause the AI system to process each input frame in a sequence of input frames through layers of a deep neural network (DNN) to produce a sequence of outputs. At least some of the layers of the DNN include a dual self-attention module having a dual non-causal and causal architecture attending to non-causal frames and causal frames. Further, the AI system renders the sequence of outputs.

Resumen de: WO2023278712A1

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a surrogate neural network configured to determine a predicted performance measure of a hardware accelerator having a target hardware configuration on a target application. The trained instance of the surrogate neural network can be used, in addition to or in place of hardware simulation, during a search process for determining hardware configurations for application-specific hardware accelerators, i.e., hardware accelerators on which one or more neural networks can be deployed to perform one or more target machine learning tasks.

Resumen de: KR20220170658A

본 개시는 제품의 이상 여부를 결정하는 방법 및 이를 수행하는 전자 장치에 관한 것이다. 일 실시 예에 의하면, 전자 장치가 인공 지능 모델을 이용하여 제품의 이상 여부를 결정하는 방법은 상기 제품에 대한 이미지를 획득하는 단계; 상기 획득된 이미지가 입력되면, 상기 제품이 정상인지 여부에 대한 검사 정보를 출력하도록, 소정의 페이크 학습 이미지에 기초하여 미리 학습되는 인공 지능 모델에, 상기 이미지를 입력함으로써, 상기 인공 지능 모델로부터 상기 검사 정보를 획득하는 단계; 및 상기 획득된 검사 정보를 출력하는 단계; 를 포함할 수 있다.

Resumen de: US2022414573A1

An initial sequence representing a partially configured engineering project is processed by a recurrent neural network to generate recommendations being a sequence of complementary items that completes an engineering project. A feature predictor component computes a set of features for each recommendation. A bisection component selects a feature from the sets of features that distinguishes some of the recommendations and forms pruned recommendations by choosing all instances from the recommendations that have the selected feature. A user interface displays the selected feature, detects a user interaction indicating that the selected feature is required, outputs the pruned recommendations. The engineering project is completed by combining the initial sequence with the chosen pruned recommendation. As a result, a user is supported in choosing optimal modules, as the selected feature can distinguish the recommendations that have the desired technical properties or target system KPI.

Resumen de: US2022414471A1

Methods and computer-readable media for repeated holdout validation include collecting independent data representing independent variables; collecting dependent data representing a dependent variable; correlating the independent data with the dependent data; creating a data set comprising the correlated independent and dependent data; generating a plurality of unique seeds; creating a plurality of training sets and a plurality of validation sets; associating each training set with a single validation set; training the neural network a plurality of times with the training sets and seeds to create a plurality of models; calculating accuracy metric values for the models using the validation sets associated with the training sets used to create respective models; performing a statistical analysis of the accuracy metric values; and ranking the independent variables by a strength of correlation of individual independent variables with the dependent variable, when a metric of the statistical analysis exceeds a threshold.