Learning from Data: Why Probability Matters
Recorded 27 October 2011 in Lausanne, Vaud, Switzerland
Event: KTN - Know Thy Neighbor
Machine learning seeks to automate the processing of large complex datasets by adaptive computing, a core strategy to meet rapidly growing demands from science and applications. In order to successfully learn from data, we literally have to give up knowing (exactly) what we are doing, but instead quantify what we don't know and reason about our uncertainties.
I will use an example from medical imaging to show how adaptive Bayesian computing and decision making from uncertain knowledge can be used in order to sequentially optimize the sampling of image bitmaps, with the goal of speeding up magnetic resonance image acquisition.
Watched 17359 times.Watch