Deep learning and weather simulations discussed at applied mathematics meeting

Leading researchers in the field of applied mathematics came together recently as Trinity played co-host to a prestigious annual meeting, with deep learning, weather simulations and computational science among the applications in the spotlight.

Kirk M Soodhalter, Ussher Assistant Professor in Trinity’s School of Mathematics, worked with Professor Jennifer Scott from the University of Reading to organise the online meeting of the Society for Industrial and Applied Mathematics (SIAM UKIE).

Trinity will host the meeting in-person next year, providing the COVID-pandemic’s impacts have lessened.

This year’s online meeting boasted talks by five plenary speakers covering a wide range of applied mathematical topics.

Dr David Barrett from Google Deepmind presented new mathematical insights to explain the surprising effectiveness of the so-called Deep Learning algorithms, while  Professor Sarah Dance of the University of Reading discussed how mathematical weather simulations can be updated real-time weather data to improve forecasts.

Professor Marco Marletta of Cardiff University discussed the feasibility of determining the electromagnetic properties of an object by taking certain measurements only at the surface of that object, while Professor Valeria Simoncini of the Università di Bologna presented some of her work on techniques for solving very large, structured sets of algebra equations which arise in many computational science problems.

Dr John Pearson of the University of Edinburgh presented general computational strategies for solving optimisation problems in scientific and industry where there is a physics-based constraint, with one goal, for example, being to minimise the financial cost of an industrial chemical process.

In addition, a number of PhD candidates delivered ten-minute talks about their work, affording them important networking opportunities early in their careers.