In one embodiment, a supervisory service for a wireless network obtains frequency-time Doppler profile information for an endpoint node attached to a first access point in the wireless network. The supervisory service uses the frequency-time Doppler profile information for the endpoint node as input to a machine learning model. The machine learning model is trained to output an action for the endpoint node with respect to the wireless network. The supervisory service causes the action for the endpoint node with respect to the wireless network to be performed.

A machine learning model including an autoencoder may be trained based on training data that includes sequences of non-anomalous performance metrics from an information technology system but excludes sequences of anomalous performance metrics. The trained machine learning model may process a sequence of performance metrics from the information technology system by generating an encoded representation of the sequence of performance metrics and generating, based on the encoded representation, a reconstruction of the sequence of performance metrics. An occurrence of the anomaly at the information technology system may be detected based on a reconstruction error present in reconstruction of the sequence of performance metrics. Related systems, methods, and articles of manufacture are provided.

A system and method for collecting data regarding operation of a robot using, at least in part, responses from a first operation model to an input of sensed data from a plurality of sensors. The collected data can be used to optimize the first operation model to generate a second operation model. While the first operation model is being optimized, a train data-driven model that utilizes an end-to-end learning approach can be generated that is based, at least in part, on the collected data. Both the second operation model and the train data-driven model can be evaluated, and, based on such evaluation, a determination can be made as to whether the train data-driven model is reliable. Moreover, based on a comparison of the models, one of the second operation model and the train data-driven model can be selected for validation, and if validated, used in the operation of the robot.

A computer-implemented method and system for calibrating an occupancy grid mapping a vehicle environment are disclosed. An example method includes identifying a feature of an occupancy grid that maps a vehicle environment in which the occupancy grid provides a primary representation of the feature. The example method also includes determining a quality level of the primary representation of the feature and determining if the quality level satisfies a quality criterion. The example method further includes adjusting a calibration of the occupancy grid if the quality level fails to satisfy the quality criterion. The adjustment of the calibration of the occupancy grid can include adjusting at least one parameter used to generate the occupancy grid to cause the quality level to satisfy the quality criterion.

A method is provided that includes generating a visual environment for interactive development of a machine learning (ML) model. The method includes accessing observations of data each of which includes values of independent variables and a dependent variable, and performing an interactive exploratory data analysis (EDA) of the values of a set of the independent variables. The method includes performing a feature construction and selection based on the interactive EDA, and in which select independent variables are selected as or transformed into a set of features for use in building a ML model to predict the dependent variable. The method includes building the ML model using a ML algorithm, the set of features, and a training set produced from the set of features and observations of the data. And the method includes outputting the ML model for deployment to predict the dependent variable for additional observations of the data.

Techniques are provided for selection of machine learning algorithms based on performance predictions by trained algorithm-specific regressors. In an embodiment, a computer derives meta-feature values from an inference dataset by, for each meta-feature, deriving a respective meta-feature value from the inference dataset. For each trainable algorithm and each regression meta-model that is respectively associated with the algorithm, a respective score is calculated by invoking the meta-model based on at least one of: a respective subset of meta-feature values, and/or hyperparameter values of a respective subset of hyperparameters of the algorithm. The algorithm(s) are selected based on the respective scores. Based on the inference dataset, the selected algorithm(s) may be invoked to obtain a result. In an embodiment, the trained regressors are distinctly configured artificial neural networks. In an embodiment, the trained regressors are contained within algorithm-specific ensembles. Techniques are also provided for optimal training of regressors and/or ensembles.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting actions to be performed by an agent interacting with an environment. In one aspect, a method comprises, at each of one or more time steps: generating a respective action score for each action in a set of possible actions, wherein the set of possible actions comprises: (i) a plurality of atomistic actions, and (ii) one or more optimization actions, wherein each optimization action is associated with a respective objective function that measures performance of the agent on a corresponding auxiliary task; selecting an action from the set of possible actions in accordance with the action scores, wherein the selected action is an optimization action; in response to selecting the optimization action, performing a numerical optimization to identify a sequence of one or more atomistic actions that are predicted to optimize the objective function.

Machine learning adversarial campaign mitigation on a computing device. The method may include deploying an original machine learning model in a model environment associated with a client device; deploying a classification monitor in the model environment to monitor classification decision outputs in the machine learning model; detecting, by the classification monitor, a campaign of adversarial classification decision outputs in the machine learning model; applying a transformation function to the machine learning model in the model environment to transform the adversarial classification decision outputs to thwart the campaign of adversarial classification decision outputs; determining a malicious attack on the client device based in part on detecting the campaign of adversarial classification decision outputs; and implementing a security action to protect the computing device against the malicious attack.

A centralized system may collect and aggregate assessments from multiple websites. An aggregate score may be calculated for the resource that cumulatively considers assessments from a plurality of different websites from which assessments are received from users. Text descriptions associated with each of the assessments may be provided to a machine-learning system that uses a trained model to assign identifiers to the assessments as they are received. These identifiers may include common words or text that are descriptive of different facets of user experiences related to receiving and using the resource. After selecting one or more identifiers, assessments associated with that identifier may be included or excluded from the display. Additionally, the overall aggregate score for the resource may be recalculated by removing components of that score that are based on assessments with identifiers that have been selected for exclusion.

The present disclosure relates to computer-implemented methods, software, and systems for predicting failure event occurrence for a machine asset. Run-to-failure sequences of time series data that include an occurrence of a failure event for the machine asset are received. One or more candidate cut-off values are determined based on iterative evaluation of a plurality of potential cut-off points. A candidate cut-off value is identified as substantially corresponding to a local peak point for calculated distances between relative frequency distributions of positive and negative sub-sequences. A failure prediction model is iteratively trained to iteratively extract sets of relevant features to determine a prediction horizon for an occurrence of the failure event for the machine asset. A candidate cut-off value associated with a model of highest quality from a set of failure prediction models determined during the iterations is selected to determine the prediction horizon for the machine asset.