Abstract: Mechanisms for performing intelligent federated machine learning (ML) model updates are provided. A plurality of ML model updates, and a plurality of dataset sketch commitment data structures (sketches), are received from a plurality of participant computing systems. Each sketch provides statistical characteristics of a corresponding local dataset used by a corresponding participant to train a local ML model. A potentially malicious participant identification operation is performed based on an analysis of the plurality of sketches to identify one or more potentially malicious participants based on differences in sketches. ML model updates received from participant computing systems identified as potentially malicious participants are discarded to thereby generate a modified set of updates. The federated ML computer model is updated based on the modified set of updates.

Abstract: A disclosed example includes implementing a first worker instance and a second worker instance to operate in parallel; running a first tuning operation via the first worker instance to tune first hyperparameters; running a second tuning operation via the second worker instance using a Bayesian-based optimization to determine a hyperparameter configuration to evaluate next; evaluating the hyperparameter configuration for an external model using a surrogate model; and selecting the hyperparameter configuration for the external model.

Abstract: Disclosed is a method for updating a node model that resists discrimination propagation in federated learning. The method includes: obtaining a node model corresponding to a data node; calculating a mean value of the distribution of class features and a quantity ratio corresponding to training data of the data node, calculating a distribution weighted aggregation model based on the node model, the mean value of the distribution of class features and the quantity ratio; calculating a regularization term corresponding to the data node based on the node model and the distribution weighted aggregation model; calculating a variance of the distribution of the class features corresponding to the data node, calculating a class balanced complementary term by using a cross-domain feature generator; and updating the node model based on the distribution weighted aggregation model, the regularization term, and the class balanced complementary term.

Abstract: The present disclosure provides a method, apparatus, electronic device and storage medium for training a semantic similarity model, which relates to the field of artificial intelligence. A specific implementation solution is as follows: obtaining a target field to be used by a semantic similarity model to be trained; calculating respective correlations between the target field and application fields corresponding to each of training datasets in known multiple training datasets; training the semantic similarity model with the training datasets in turn, according to the respective correlations between the target field and the application fields corresponding to each of the training datasets.

Abstract: A data processing system includes compile time logic to section a graph into a sequence of sections, including a first section followed by a second section. The compile time logic configured the first section to generate a first output in a first non-overlapping target configuration in response to processing an input in a first overlapping input configuration, and configures the second section to generate a second output in a second non-overlapping target configuration in response to processing the first output in a second overlapping input configuration. The compile time logic also creates a set of computer instructions to execute the first section and the second section on a target processing system.

Abstract: The disclosure discloses a method and an apparatus for generating a semantic representation model, and a storage medium. The detailed implementation includes: performing recognition and segmentation on the original text included in an original text set to obtain knowledge units and non-knowledge units in the original text; performing knowledge unit-level disorder processing on the knowledge units and the non-knowledge units in the original text to obtain a disorder text; generating a training text set based on the character attribute of each character in the disorder text; and training an initial semantic representation model by employing the training text set to generate the semantic representation model.

Abstract: A data processing method and apparatus using a neural network, and an electronic device including the data processing apparatus. The data processing method includes identifying an operator that selects one of a plurality of execution paths for a portion of the neural network while sequentially executing layers included in the neural network, selecting a specific execution path, from among the plurality of execution paths, based on a remaining time that is left for an inference of the neural network, and obtaining a result of the inference of the neural network through the specific execution path.

Abstract: Disclosed is a method for distributing work points to a plurality of task-performing robots, the method performed by one or more processors of a computing device. The method may include: determining available work points for each of the plurality of task-performing robots; distributing target work points for each of the plurality of task-performing robots based on the determined available work points, and predicting a plurality of target work trajectories for each of the plurality of task-performing robots based on the distributed target work points; calculating a first loss or a second loss based on the predicted plurality of target work trajectories; and re-distributing target work points to each of the plurality of task-performing robots based on at least one of the calculated first loss or the calculated second loss.

Abstract: Systems, methods, and computer readable media related to information retrieval. Some implementations are related to training and/or using a relevance model for information retrieval. The relevance model includes an input neural network model and a subsequent content neural network model. The input neural network model and the subsequent content neural network model can be separate, but trained and/or used cooperatively. The input neural network model and the subsequent content neural network model can be “separate” in that separate inputs are applied to the neural network models, and each of the neural network models is used to generate its own feature vector based on its applied input. A comparison of the feature vectors generated based on the separate network models can then be performed, where the comparison indicates relevance of the input applied to the input neural network model to the separate input applied to the subsequent content neural network model.

Abstract: A system and method includes generating approximate distributions for distinct classes of data samples; computing a first partial Jensen-Shannon (JS) divergence and a second partial JS divergence based on the approximate distribution of the disparity affected class of data samples with reference to the approximate distribution of the control class of data samples; computing a disparity divergence based on the first partial JS divergence and the second partial JS divergence; generating a distribution-matching term based on the disparity divergence, wherein the distribution-matching term mitigates an inferential disparity between the control class of data samples and the disparity affected class of data samples during a training of an unconstrained artificial neural network; constructing a disparity-constrained loss function based on augmenting a target loss function with the distribution-matching term; and transforming the unconstrained ANN to a disparity-constrained ANN based on a training of the unconstraine

Abstract: A plurality of initial machine learning models are determined based on a plurality of original features. The plurality of initial machine learning models are filtered by selecting a subset of the initial machine learning models as one or more surviving machine learning models. One or more evolved machine learning models are generated. At least one of the evolved machine learning models is based at least in part on one or more new features, which are based at least in part on a transformation of at least one of features of the one or more surviving machine learning models. Corresponding validation scores associated with the one or more evolved machine learning models and corresponding validation scores associated with the one or more surviving machine learning models are compared. At least one of the one or more evolved machine learning models or the one or more surviving machine learning models are selected as one or more new selected surviving machine learning models.

Abstract: A method includes installing, in a consumer database account, a shared-instance database that includes a shared instance of a provider-account database that resides in a provider database account. The shared-instance database includes a first schema that includes provider-account training data, provider-account scoring data, a training function, and a scoring function. The method also includes invoking the training function from the consumer database account, which results in creation in the consumer database account of a second schema that includes a machine-learning-model instance of a machine learning model, and which also results in training the machine-learning model instance with at least the provider-account training data. Additionally, the method includes generating consumer-account scoring data by inputting, into the trained machine-learning-model instance, consumer-account input data that is stored in the consumer database account.

Abstract: Disclosed are a data processing method and apparatus based on an optical neural network, a computer-readable storage medium, and an optical neural network.

Abstract: A processing graph of an application with a sequence of processing nodes is obtained which processes an input and generates an intermediate representation a further intermediate representation, and an output representation of the input at stages in the sequence of processing nodes. Graph metadata is generated that specifies a non-overlapping target tiling configuration for the output representation, an overlapping tiling configuration for the input, an overlapping tiling configuration for the intermediate representation, and a third tiling configuration for the further intermediate representation. The processing graph is modified based on the graph metadata to conform to the parameters specified by the graph metadata. A set of computer instructions is then created to execute the modified processing graph on a target processing system.

Abstract: Disclosed is a data processing system that includes compile time logic to section a graph into a sequence of sections including a first section and a second section. The compile time logic is to configure the first section with a first topology of tiling configurations in which to tile inputs, intermediate outputs, and final outputs of the first section, and configure the second section with a second topology of tiling configurations in which to tile inputs, intermediate outputs, and final outputs of the second section. The data processing system further includes runtime logic configured with the compile time logic to execute the first section to generate the inputs, intermediate outputs, and final outputs of the first section in the first topology of tiling configurations, and execute the second section to generate the inputs, intermediate outputs, and final outputs of the second section in the second topology of tiling configurations.

Abstract: A method for training a neural network model having layers and parameters, comprises providing an input corresponding to each of samples comprised in a batch of a training dataset to an input layer, obtaining outputs from the neural network model, calculating a loss function for each of the samples based on the outputs and corresponding desired values, and determining values for minimizing a mismatch between the outputs and the corresponding desired values across the samples for the parameters based on the loss function. The determining of the values comprises running a forward model through the layers, determining a Sifrian functional for the layers based on the forward model, backpropgation, and gradient definition, determining equations from the Sifrian functional, performing transformation and optionally reduction to get a pivotal Sifr equation and solving the equation to estimate a second-order update for the purpose of the neural network model training.

Abstract: The present disclosure provides systems and methods for on-device machine learning. In particular, the present disclosure is directed to an on-device machine learning platform and associated techniques that enable on-device prediction, training, example collection, and/or other machine learning tasks or functionality. The on-device machine learning platform can include a mode controller that allows a user to provide data input indicating whether to operate one or more applications on the device in a first collection mode (e.g., permission mode) for storing training examples or a second collection mode for (e.g., incognito mode) for not storing training examples. The training examples can be generated based on user interaction with the one or more applications and used to personalize one or more machine-learned models used by the application(s) by retraining the models using the user-specific training examples.

Abstract: A device may receive log data from application logs associated with applications, service logs associated with services, and server logs associated with server devices. The device may store the log data. The device may perform natural language processing on the log data to convert the log data into event data identifying events associated with categories. The device may process the event data, with a first machine learning model, to identify patterns in the event data and to generate an alert based on the patterns. The device may process the event data, with a second machine learning model, to generate a correlation matrix for the event data and to predict an event based on the correlation matrix. The device may process the event data, with a third machine learning model, to classify the event data based on the categories and to generate a recommendation based on classifying the event data.

Abstract: A system and method includes generating approximate distributions for distinct classes of data samples; computing a first partial Jensen-Shannon (JS) divergence and a second partial JS divergence based on the approximate distribution of the disparity affected class of data samples with reference to the approximate distribution of the control class of data samples; computing a disparity divergence based on the first partial JS divergence and the second partial JS divergence; generating a distribution-matching term based on the disparity divergence, wherein the distribution-matching term mitigates an inferential disparity between the control class of data samples and the disparity affected class of data samples during a training of an unconstrained artificial neural network; constructing a disparity-constrained loss function based on augmenting a target loss function with the distribution-matching term; and transforming the unconstrained ANN to a disparity-constrained ANN based on a training of the unconstraine

Abstract: A reasoning engine is disclosed. Contemplated reasoning engines acquire data relating to one or more aspects of various environments. Inference engines within the reasoning engines review the acquire data, historical or current, to generate one or more hypotheses about how the aspects of the environments might be correlated, if at all. The reasoning engine can attempt to validate the hypotheses through controlling acquisition of the environment data.