Description

Finding Shark Attacks in Data Streams (Using Machine Learning to Detect Anomalies)

In this lecture, Sebastian Steinke, a software engineer at KAUST who started to learn Machine Learning via self-directed learning, will introduce a model that is able to predict future observations with high accuracy and detects incoming data that doesn't fit.

The Integrated Ocean Processes (IOP) group uses Seaglider autonomous underwater vehicles to conduct oceanographic sampling in long-duration missions that can last several months. Gliders dive up to a depth of 1,000 m while sensors on board constantly measure environmental parameters. Using autonomous operating Gliders helps scientists to sample on long missions and in all weather conditions.

Unfortunately, throwing your hardware into the ocean also comes with risks. Incidents like collisions with boats, hardware failures or even shark attacks can led to the loss of a Glider. Therefore, it is important to identify such events as soon as possible to organize recovery.

To help human pilots interpreting incoming data from the Glider fleet and recognize anomalies as fast as possible, they are training models on incoming data streams.

We value your feedback and welcome any comments you may have to help us improve our programs. Make sure you submit the survey after you have attended this event!