Token-wise training for attention based end-to-end speech recognition

    公开(公告)号:US11636848B2

    公开(公告)日:2023-04-25

    申请号:US17316856

    申请日:2021-05-11

    摘要: A method of attention-based end-to-end (A-E2E) automatic speech recognition (ASR) training, includes performing cross-entropy training of a model, based on one or more input features of a speech signal, determining a posterior probability vector at a time of a first wrong token among one or more output tokens of the model of which the cross-entropy training is performed, and determining a loss of the first wrong token at the time, based on the determined posterior probability vector. The method further includes determining a total loss of a training set of the model of which the cross-entropy training is performed, based on the determined loss of the first wrong token, and updating the model of which the cross-entropy training is performed, based on the determined total loss of the training set.

    Electronic logging of vehicle and driver data for compliance support and prediction

    公开(公告)号:US11636369B2

    公开(公告)日:2023-04-25

    申请号:US16557437

    申请日:2019-08-30

    摘要: A method for electronic logging of carrier data is described. The logged electronic carrier data is utilized to predict a likelihood of occurrence of a potential impending compliance rule violation. A plurality of travel conditions capable of changing a transit period, such as weather, traffic, and construction are monitored, aggregated, and used to determine a predicted likelihood of changing the transit period. The likelihood of occurrence of a compliance rule violation is determined, based on the aggregated values, the predicted likelihood of changing the transit period, and a group of compliance rules. An indicator of the likelihood of occurrence of a compliance rule violation may be generated, and presented to any number of electronic devices. Remedial actions to avoid or mitigate the compliance rule violation may be suggested.

    SYSTEMS AND METHODS FOR MISSING DATA IMPUTATION

    公开(公告)号:US20230121411A1

    公开(公告)日:2023-04-20

    申请号:US17949076

    申请日:2022-09-20

    摘要: Congestive heart failure (CHF) is a leading cause of death in the United States. WANDA is a wireless health project that leverages sensor technology and wireless communication to monitor the health status of patients with CHF. The first pilot study of WANDA showed the system’s effectiveness for patients with CHF. However, WANDA experienced a considerable amount of missing data due to system misuse, nonuse, and failure. Missing data is highly undesirable as automated alarms may fail to notify healthcare professionals of potentially dangerous patient conditions. Embodiments of the present disclosure may utilize machine learning techniques including projection adjustment by contribution estimation regression (PACE), Bayesian methods, and voting feature interval (VFI) algorithms to predict both non-binomial and binomial data. The experimental results show that the aforementioned algorithms are superior to other methods with high accuracy and recall.

    PREDICTION OF PEPTIDE CLEAVAGE IN POLYPEPTIDES THROUGH PHYSICS-BASED SIMULATIONS

    公开(公告)号:US20230120018A1

    公开(公告)日:2023-04-20

    申请号:US18058110

    申请日:2022-11-22

    申请人: Genentech, Inc.

    IPC分类号: G16B15/20 G16B40/20 G06N7/01

    摘要: The present disclosure relates to polypeptide degradation, and in particular to techniques for predicting the likelihood that a peptide bond for a given polypeptide molecule is susceptible to a cleavage reaction. Particularly, aspects of the present disclosure are directed to generating a representation of a polypeptide, performing a molecular-dynamics simulation using the representation to obtain a set of polypeptide conformations, determining, for each polypeptide conformation, a spatial characteristic of an amino acid, estimating a nucleophilic attack distance of each polypeptide conformation based on the spatial characteristic, identifying a reactive conformation that is susceptible to a cleavage reaction based on the nucleophilic attack distance of each polypeptide conformation, determining a free energy of the spatial characteristic of the amino acid in the reactive conformation; and predicting a probability of the side chain of the amino acid being trapped in the reactive conformation based on the free energy.

    TECHNOLOGIES FOR PREOPERATIVE IMPLANT SIZE ESTIMATION

    公开(公告)号:US20230111368A1

    公开(公告)日:2023-04-13

    申请号:US18079596

    申请日:2022-12-12

    摘要: A computing system including a processor and memory comprising a plurality of instructions stored thereon that, in response to execution by the processor, causes the computing system to determine a plurality of implant size predictions with associated confidence levels based on one or more patient or surgical parameters, wherein each of the implant size predictions identifies a confidence level that a prospective implant of a corresponding size will fit a patient, determine whether a combined confidence level determined based on a subset of the plurality of associated confidence levels is at least a threshold value, and recommend, in response to a determination that the combined confidence level is not at least the threshold value, incorporation of at least one of an additional implant size prediction of the plurality of implant size predictions in the subset or digital templating data to improve an accuracy of an implant size estimation.

    INFORMATION PROCESSING METHOD AND RELATED DEVICE

    公开(公告)号:US20230087821A1

    公开(公告)日:2023-03-23

    申请号:US18071316

    申请日:2022-11-29

    摘要: This application disclose an information processing method and a related device. This application provide a first AI entity in an access network, and define a plurality of basic interaction modes between the first AI entity and a terminal device. In an interaction mode, the first AI entity may receive second AI model information sent by the terminal device. The second AI model information does not include user data of the terminal device. The first AI entity may update first AI model information of the first AI entity based on the second AI model information, and then send updated first AI model information to the terminal device, so that the terminal device trains and updates the second AI model information.

    DIGITAL HISTOPATHOLOGY AND MICRODISSECTION

    公开(公告)号:US20230129222A1

    公开(公告)日:2023-04-27

    申请号:US18088487

    申请日:2022-12-23

    申请人: NantOmics, LLC

    发明人: Bing Song Gregory Chu

    摘要: A computer implemented method of generating at least one shape of a region of interest in a digital image is provided. The method includes obtaining, by an image processing engine, access to a digital tissue image of a biological sample; tiling, by the image processing engine, the digital tissue image into a collection of image patches; identifying, by the image processing engine, a set of target tissue patches from the collection of image patches as a function of pixel content within the collection of image patches; assigning, by the image processing engine, each target tissue patch of the set of target tissue patches an initial class probability score indicating a probability that the target tissue patch falls within a class of interest, the initial class probability score generated by a trained classifier executed on each target tissue patch; generating, by the image processing engine, a first set of tissue region seed patches by identifying target tissue patches having initial class probability scores that satisfy a first seed region criteria, the first set of tissue region seed patches comprising a subset of the set of target tissue patches; generating, by the image processing engine, a second set of tissue region seed patches by identifying target tissue patches having initial class probability scores that satisfy a second seed region criteria, the second set of tissue region seed patches comprising a subset of the set of target tissue patches; calculating, by the image processing engine, a region of interest score for each patch in the second set of tissue region seed patches as a function of initial class probability scores of neighboring patches of the second set of tissue region seed patches and a distance to patches within the first set of issue region seed patches; and generating, by the image processing engine, one or more region of interest shapes by grouping neighboring patches based on their region of interest scores.

    System and Method to Facilitate Welding Software as a Service

    公开(公告)号:US20230129188A1

    公开(公告)日:2023-04-27

    申请号:US17828630

    申请日:2022-05-31

    发明人: Christopher Hsu

    摘要: A weld production knowledge system for processing welding data collected from one of a plurality of welding systems, the weld production knowledge system comprising a communication interface communicatively coupled with a plurality of welding systems situated at one or more physical locations. The communication interface may be configured to receive, from one of said plurality of welding systems, welding data associated with a weld. The weld production knowledge system may comprise an analytics computing platform operatively coupled with the communication interface and a weld data store. The weld data store employs a dataset comprising (1) welding process data associated with said one or more physical locations, and/or (2) weld quality data associated with said one or more physical locations. The analytics computing platform may employ a weld production knowledge machine learning algorithm to analyze the welding data vis-à-vis the weld data store to identify a defect in said weld.