Methods and systems are disclosed for generating optimal data predictions in time series data signals based on empirically-optimized model selection, noise filtering, and window size selection using machine learning models. For example, the system may receive a first subset of time series data. The system may receive a prediction horizon. The system may generate a feature input based on the first subset of time series data and the prediction horizon. The system may input the feature input into a machine learning model, wherein the machine learning model includes multiple components. The system may receive an output from the machine learning model. The system may generate for display, on a user interface, a prediction for the first subset of time series data at the prediction horizon based on the output.

Systems and methods are provided for detecting a flashing light on one or more traffic signal devices. The method includes capturing a series of images of one or more traffic signal elements in a traffic signal device over a length of time. The method further includes, for each traffic signal element, analyzing the series of images to determine one or more time periods at which the traffic signal element is in an on state or an off state, and analyzing the time periods to determine one or more distinct on states and one or more distinct off states. The method further includes identifying one or more cycles correlating to a distinct on state immediately followed by a distinct off state, or a distinct off state immediately followed by a distinct on state, and, upon identifying a threshold number adjacent cycles, classifying the traffic signal element as a flashing light.

A data prediction subsystem includes receives event data indicating amounts of items removed from locations over a previous period of time. For a first day of the first set of event data having zero events or an empty status indicating that the first item is not believed to be present at the first location, longitudinal and cross-sectional components are determined. An anticipated event value for the first item at the first location is determined using the longitudinal component and the cross-sectional component. Based at least in part on the anticipated event value, a prediction value is determined that corresponds to a recommended amount of the first item to request at a future time.

Methods, systems, and computer-readable storage media for receiving a first set of predictions generated by a ML model during execution of a training pipeline to train the ML model, each prediction in the first set of predictions being associated with a confidence, determining a set of confidence bins based on confidences of the first set of predictions, for each confidence bin in the set of confidence bins, providing an accuracy, processing the set of confidence bins and accuracies through a regression model to provide one or more regressions, each regression representing a confidence-to-accuracy relationship, defining a set of confidence thresholds based on at least one regression of the one or more regressions, and during an inference phase, applying the set of confidence thresholds to selectively filter predictions from a second set of predictions generated by the ML model.

According to one embodiment, a learning apparatus includes processing circuitry. The processing circuitry generates a plurality of pieces of partial data from a mini-batch of learning data used for a plurality of learning processes for learning of a parameter of a neural network using an objective function, calculates a partial gradient that is a gradient related to the parameter of the objective function for each of the pieces of partial data, and updates the parameter based on an average value of the plurality of partial gradients corresponding to the pieces of partial data and a variance for the partial gradients.

A method for routing, through a geographic region over a time interval, a sensor platform mounted on a vehicle is described. The method includes receiving a precision level for at least one constituent of an environment measured by a sensor of the sensor platform. The precision level corresponds to a mean concentration of the constituent(s) over the time interval. A reference dataset corresponding to the geographic region and the time interval is selected. From the reference dataset and the precision level, at least one minimum number of distinct samples for a plurality of geographic segments of the geographic region is determined. The method also includes determining a number of passes for the geographic region over the time interval using the minimum number of distinct samples for each of the plurality of geographic segments. Each pass of the number of passes is part of a route for the vehicle.

A dividing pattern determination device capable of reducing the amount of computation performed when determining a dividing pattern of an image. An image for which a dividing pattern is expressed by a hierarchical structure for each predetermined area is input to a feature extraction section, and the feature extraction section generates, based on the input image, for the predetermined area, a hierarchy map in which a value indicative of a block size is associated with each of a plurality of blocks in the predetermined area. A determination section determines a dividing pattern of the image based on the generated hierarchy map.

A system and method for machine learning features validation. A method includes: performing statistical testing on a plurality of pairs of features, each pair of features including a test feature of a plurality of test features extracted from a first data set and a corresponding training feature extracted from a second data set during a training phase for a machine learning model, wherein the statistical testing is performed under a null hypothesis that the first data set and the second data set are drawn from a same continuous distribution, wherein performing the statistical testing further comprises determining a degree to which each test feature of the plurality of pairs of features deviates from the corresponding training feature; and determining, based on the degree to which each test feature of the plurality of pairs of features deviates from the corresponding training feature, whether the plurality of test features is validated.

A method for generating a facial animation from a single image is provided. The method is mainly divided into four steps: generation of facial feature points in an image, global two-dimensional deformation of the image, optimization of details of a facial area, and generation of texture of an oral cavity area. The present disclosure can generate a facial animation in real time according to a change of the facial feature points, and the animation quality reaches a level of current state-of-art facial image animation technology. The present disclosure can be used in a series of applications, such as facial image editing, portrait animation generation based on a single image, and facial expression editing in videos.

An adaptive processing method for new scenes in autonomous driving, comprising: obtaining scene data corresponding to new scene of vehicle driving, wherein the scene data describes vehicles state and driving operations in the new scene; obtaining a test set of the new scene based on processing the scene data by a preset distribution; updating parameters of a pre-training model by inputting the test set, and obtaining a scene model adapted to the new scene based on gradient iteration of general model parameters of the pre-training model, wherein the scene model is configured to output an autonomous driving strategy for the vehicle in the new scene. Therefore, the autonomous driving vehicle transforms a new scene to a known scene, and no longer be troubled by unpredictable new scenes, and greatly enhance the reliability and stability of autonomous driving.