Abstract: A neural net processor provides twin processing paths trainable using different moments of the input data, one moment providing a proxy for uncertainty. Subsequent operation of the trained neural net allows monitoring of the uncertainty proxy to provide real-time assessment of neural net model-based uncertainty.

Abstract: A MapReduce-based training framework exploits both data parallelism and model parallelism to scale training of complex models. Particular model architectures facilitate and benefit from use of such training framework. As one example, a machine-learned model can include a shared feature extraction portion configured to receive and process a data input to produce an intermediate feature representation and a plurality of prediction heads that are configured to receive and process the intermediate feature representation to respectively produce a plurality of predictions. For example, the data input can be a video and the plurality of predictions can be a plurality of classifications for content of the video (e.g., relative to a plurality of classes).

Abstract: A MapReduce-based training framework exploits both data parallelism and model parallelism to scale training of complex models. Particular model architectures facilitate and benefit from use of such training framework. As one example, a machine-learned model can include a shared feature extraction portion configured to receive and process a data input to produce an intermediate feature representation and a plurality of prediction heads that are configured to receive and process the intermediate feature representation to respectively produce a plurality of predictions. For example, the data input can be a video and the plurality of predictions can be a plurality of classifications for content of the video (e.g., relative to a plurality of classes).

Abstract: A method of improving data sets, for example, of patients, each being characterized by relatively low-cost medical data, identifies those patients where the acquisition of higher cost medical data would best inform an estimate of the higher cost medical data for the remaining patients. In this way scarce medical resources can be more efficiently applied in characterizing a potential patient pool, for example, for a clinical trial when resources are not available for extensive medical characterization of each trial participant.

Abstract: A neural net processor provides twin processing paths trainable using different moments of the input data, one moment providing a proxy for uncertainty. Subsequent operation of the trained neural net allows monitoring of the uncertainty proxy to provide real-time assessment of neural net model-based uncertainty.

Abstract: A method of improving data sets, for example, of patients, each being characterized by relatively low-cost medical data, identifies those patients where the acquisition of higher cost medical data would best inform an estimate of the higher cost medical data for the remaining patients. In this way scarce medical resources can be more efficiently applied in characterizing a potential patient pool, for example, for a clinical trial when resources are not available for extensive medical characterization of each trial participant.

Abstract: A solution is provided for object tracking in a sports video is disclosed. A determination whether a position of the object was identified in a previous video frame is made. If the position of the object was identified in the previous video frame, a new position of the object is identified in a current video frame based on the identified position of the object in the previous video frame. An expected position of the object is identified based on the identified position of the object in the previous video frame and a trained object classification model. A determination is made whether the new position is consistent with the expected position; if the new position is consistent with the expected position, the new position is stored as the position of the object in the current frame.

Abstract: A solution is provided for object tracking in a sports video is disclosed. A determination whether a position of the object was identified in a previous video frame is made. If the position of the object was identified in the previous video frame, a new position of the object is identified in a current video frame based on the identified position of the object in the previous video frame. An expected position of the object is identified based on the identified position of the object in the previous video frame and a trained object classification model. A determination is made whether the new position is consistent with the expected position; if the new position is consistent with the expected position, the new position is stored as the position of the object in the current frame.