The subject technology detects anomalies in media campaign configuration settings. The anomaly detection system may leverage one or more deep learning models to detect anomalies and identify particular configuration settings that contribute to the detected anomalies. In various embodiments, two or more of the deep learning models may be combined into an ensemble model that boosts the accuracy of anomaly predictions made by the anomaly detection system. The anomaly detection system may review the configuration settings of media campaigns during the configuration process and before the media campaigns run on a publication system in order to reduce the amount of unsuccessful campaigns and minimize the amount of wasted resources spent on running campaigns that have a low likelihood of achieving user defined goals.

A machine learning model is trained with a training dataset, where the machine learning model comprises a plurality of layers. During training, values of a plurality of coefficients of one or more layers are monitored. In response to detecting a change of a given coefficient by more than a threshold during a given training run, a given reference to a given input dataset of the given training run is stored. In response to detecting an output error of a trained version of the machine learning model, the given reference to the given input dataset is retrieved if the given coefficient is located on a backward path providing more than a threshold contribution to the output error. Next, the given reference is provided to an application analyzing the trained version of the machine learning model in order to determine a cause of the output error.

The present disclosure relates to systems, non-transitory computer-readable media, and methods for selecting machine-learning models and hardware environments for executing a task. In particular, in one or more embodiments, the disclosed systems select a designated machine-learning model for executing a task based on workload features of the task and task routing metrics for a plurality of machine-learning models. In addition, in one or more embodiments, the disclosed systems select a designated hardware environment for executing the task based on workload features for the task and task routing metrics for a plurality of hardware environments. In some embodiments, the disclosed systems select a fallback machine-learning model and a fallback hardware environment for executing the task if the designated machine-learning model or designated hardware environment are unavailable. Moreover, in one or more embodiments, the disclosed systems can pause and initiate tasks based on bandwidth availability.

The present disclosure relates to systems, non-transitory computer-readable media, and methods for selecting machine-learning models and hardware environments for executing a task. In particular, in one or more embodiments, the disclosed systems select a designated machine-learning model for executing a task based on workload features of the task and task routing metrics for a plurality of machine-learning models. In addition, in one or more embodiments, the disclosed systems select a designated hardware environment for executing the task based on workload features for the task and task routing metrics for a plurality of hardware environments. In some embodiments, the disclosed systems select a fallback machine-learning model and a fallback hardware environment for executing the task if the designated machine-learning model or designated hardware environment are unavailable. Moreover, in one or more embodiments, the disclosed systems can pause and initiate tasks based on bandwidth availability.

The subject technology detects anomalies in media campaign configuration settings. The anomaly detection system may leverage one or more deep learning models to detect anomalies and identify particular configuration settings that contribute to the detected anomalies. In various embodiments, two or more of the deep learning models may be combined into an ensemble model that boosts the accuracy of anomaly predictions made by the anomaly detection system. The anomaly detection system may review the configuration settings of media campaigns during the configuration process and before the media campaigns run on a publication system in order to reduce the amount of unsuccessful campaigns and minimize the amount of wasted resources spent on running campaigns that have a low likelihood of achieving user defined goals.

A method for real time analysis, the method includes producing, by a classification unit having a neural network, classification decisions for sensed information units obtained in an environment of a vehicle; automatically generating, by one or more computing devices and having auto-labeling capabilities running in a real-time driving of the vehicle, an automated ground truth labeling for a determined set of sensed information units; detecting, by the one or more computing devices and based on a performance indication related to the automated ground truth labeling, an issue with respect to the classification detection; and responsive to the detecting, addressing the detected issue in the real-time driving of the vehicle, using a signature associated with at least the classification decision or the detected issue. The neural network is in a same state in the producing the classification decision, detecting the issue, and addressing the issue.

A method for monitoring a prediction error during the inference of an application machine learning model providing predictions based on at least one actual time-series signal from an actual sensor. The method includes: predicting an expected time-series signal from the actual time-series signal; calculating an error based on the expected signal and the actual signal; determining the a stationarity of the error; and determining the an evolution of the stationarity.

The disclosure provides misclassification-driven training (MDT) that efficiently identifies critical faults in neuromorphic hardware, such as a memristor crossbar. MDT advantageously identifies whether a hardware fault is a critical fault and can be used to limit fault recovery when a hardware fault is not a critical fault. By applying fault-tolerant techniques directed to critical faults, such as only for critical faults, processing overhead of a neural network can be reduced. In one aspect, the disclosure provides a method of identifying critical faults in neuromorphic hardware of a neural network. In one example the method of identifying includes: (1) determining a significant parameter of a trained neural network that impacts classification of a sample of a dataset, (2) obtaining a location of the significant parameter in the neuromorphic hardware, and (3) identifying the location as a critical fault of the neuromorphic hardware.

A device and method for machine learning using an artificial neural network. For a calculation hardware for the artificial neural network, a layer description is provided, which defines at least one part of a layer of the artificial neural network, the layer description defining a tensor for input values of at least one part of this layer, a tensor for weights of at least one part of this layer, and a tensor for output values of at least one part of this layer, in particular of its starting address. A message that includes a start address of the tensor for the input values, or of the tensor for the weighs, or of the tensor for the output values is sent by the calculation hardware for transfer of the input values, or the weights, or the output values, is sent by the calculation hardware.

A machine learning model is trained with a training dataset, where the machine learning model comprises a plurality of layers. During training, values of a plurality of coefficients of one or more layers are monitored. In response to detecting a change of a given coefficient by more than a threshold during a given training run, a given reference to a given input dataset of the given training run is stored. In response to detecting an output error of a trained version of the machine learning model, the given reference to the given input dataset is retrieved if the given coefficient is located on a backward path providing more than a threshold contribution to the output error. Next, the given reference is provided to an application analyzing the trained version of the machine learning model in order to determine a cause of the output error.