Gaussian Processes for Monitoring Patients with Mobile Sensors

As wearable physiological sensors become morecommon, there is a need for algorithms that can use the resultingwaveforms to perform robust data analysis. Existing techniqueshave failed to penetrate into clinical practice due to theirperceived lack of robustness. This presentation will argue thatthe natural framework for inference with noisy, incomplete datais that of Bayesian Gaussian processes. We describe the use ofmulti-task algorithms for monitoring patients via wearablesensors. Such algorithms must be able to cope with differingsampling rates between sensors; different modalities of dataacquisition; and phase offsets between sensors.