All first-year Science courses will be delivered virtually for Winter 2021. Due to the COVID-19 provincial lock-down, all students who were expecting in-person labs, should check their McMaster e-mail for important winter announcements..

Colloquium - Alex Bihlo - Deep learning in the atmospheric sciences


Title: Deep learning in the atmospheric sciences 

Speaker: Alex Bihlo (Memorial University) 

Abstract: As most fields of science, also meteorology has seen a steady increase in the interest and application of deep learning over the past two years. While still not quite comparable in terms of accuracy to many more traditional numerical techniques, initial breakthroughs in problems related to weather nowcasting or parameterization of unresolved physical processes give evidence of the potential that deep learning holds for the atmospheric sciences. 

In this talk I will present a few select applications of deep learning in meteorology. I will present a conditional generative adversarial network for local and global (ensemble) weather forecasting, an autoencoder-type precipitation nowcasting system based on two-dimensional radar reflectivity maps, as well as a local model for estimating the visibility at airports using a WaveNet-type neural network architecture.
This is joint work with Johannes Sachsperger at AustroControl, and James Jackaman and Wasiim Kausmally at Memorial University.

Location: Virtual

Join Zoom Meeting

Meeting details to follow

Go Back
McMaster University - Faculty of Science | Math & Stats