From May 13th to May 15th 2019, the Aarhus University organized a workshop “Machine Learning for Biologists” in idyllic Rønde, Denmark.
Machine learning is revolutionizing both the interpretation and the collection of citizen science data. Access to the amounts of data and computing power needed is now commonplace, and recent improvements in data handling tools make this exciting field accessible to non-computer scientists. The Aarhus University invited to a workshop sharing their experiences with the topic, especially tailored for biologists. This is a perfect match for Jan and Wouter, who are each using these technologies in their research in different ways.
Kicking off the workshop was senior researcher at Aarhus University Toke Thomas Høye in an evening session. Toke Thomas gave an introduction to the theme using an image dataset in his research on moths in the arctic. It proved really useful to start off the workshop with a real feel of what the work entails.
The next day, postdoc researcher Xiaodong Duan took us through the concepts behind the technology, introducing deep neural networks as a concept, as well as the use workflow of the popular Tensorflow API. These sessions were alternated with more hands on sessions in which researchers Hjalte Jensen and Jesper Erenskjold Moeslund, helping us to set up the tools and data to get an insight into the work they have been doing, to create input for a model that can detect flowers, their phenological stage and plant distribution modelling, respectively.
The workshop concluded with a deeper dive into the different kinds of neural networks by Xiaodong Duan, and another hands on session on the validation of species observations. After that there was a great group discussion where all participants had the opportunity to share their experiences and questions regarding their own research. It is hard to overstate how educational it is to do this as a group, receiving a lot of new ideas listening to feedback on one’s own project or suggesting ways for others to solve certain issues.
We think there could not have been a more tailor-made workshop for our research, and the alternating sessions of theoretical framework and hands on, real life projects was really helpful. In these beautiful surroundings we found time every day to do some citizen science ourselves, gaining more insight in the nature of the data we work with!