This NMR masterclass combines lectures, practical hands-on exercises and discussion, enabling valuable interaction between students and instructors.
Wednesday 21st February
09.15 - 09.45: Arrival, refreshments, software testing
09.45 - 10.00: Welcome & Introductions
10.00 - 12.45: Machine learning in NMR: how to build a training database. (Instructor: Prof Ilya Kuprov, University of Southampton)
The biggest problem with machine learning in magnetic resonance is training databases. There will never be enough NMR or EPR data on Earth to train even the smallest neural network to convergence. Therefore, training databases are usually synthetic. This workshop gives a few examples.
12.45 - 13.30: Lunch
13.30 - 16.30: Assessing the use of deep learning to reconstruct sparsely sampled spectra. (Instructor: Prof D. Flemming Hansen, University College London), **includes a 15 min refreshment break at 3pm
We will initially start with a brief lecture, where the recent developments of deep learning tools and deep neural networks (DNNs) to analyse and transform NMR data will be discussed. Subsequently there will be a practical session, where the recent FID-Net DNN will be used to reconstruct sparsely sampled 2D NMR data. Our focus will be on accessing the quality of the reconstructions and judge the limitations of reconstructions with DNNs.
Before the workshop, please make sure that you have created an account on NMRbox (nmrbox.nmrhub.org/).
16.30 - 16.45: Closing remarks & departure