Southampton Marine & Maritime Institute reposted this
Very excited to have this published from a very successful, in my view, KTP with Silverstream Technologies, based on Amy's excellent PhD research. I'll be discussing how this work helps reduce ship emissions as part of my keynote at the Singapore Maritime Forum on Thursday 7th for those of you able to join. Thanks to David Connolly for getting this all started and Terry Corner for the support through the KTP.
Our newest paper "Robust approximation of the conditional mean for applications of Machine Learning" illustrates how limiting the current set of machine learning loss functions are, and how we should be considering a wider range of (non-differentiable) loss functions in all our machine learning. As an example of this phenomena, we model ship powering on a vessel using air lubrication and find the results that fit the ground truth the best are using non-traditional loss functions. Adam Sobey Dominic Hudson Josef Camilleri https://lnkd.in/eTc_aNem