Can our toolbox benefit from “Machine Learning” ?
Machine Learning and AI have become modern buzzwords. The success (or lack thereof) of its application for a broad range of fields has raised both interest and skepticism. After an overview on the basic concepts behind machine learning algorithms, we will focus on two main applications:
- predicting storm surges and in general sea water level residual at different gauges around the world based on an ensemble of simple neural networks
- emulating ocean wave characteristics for the North Atlantic ocean using a convolutional neural network
Once trained, those algorithms are computationally-cheap and open new ways to assess uncertainty or improve model parameterisations. We will finally discuss capability and limitations of such techniques and debate how they can work together with our current tools.