idalab goes PyData
This year’s PyData Berlin conference, taking place from July 6th to 8th at the Charité-Campus (Virchow), was a huge success and with almost 600 participants the biggest PyData conference all over Europe. Concerned with providing a forum for python users and developers in the field of data analysis, a wide range of topics was covered in four simultaneous tracks, going from deep learning and scalability over data privacy and best practices to putting machine learning into production. idalab was strongly represented at the event: I was part of the organizing team and my colleagues Erik Schumacher, David Hinrichs and Hagen Hoferichter (f.l.t.r. in the above image) joined the conference as attendees.
Focussing especially on ethics in data science, we had several interesting keynotes around GDPR by Marit Hansen, hacking the iron curtain by Andrada Fiscutean as well as fairness in online social systems by Elisa Celis. Additionally, we at PyData put a lot of effort into supporting people of diverse background to find their way into the data science community and were thus very happy to have Nakeema Stefflbauer – the founder of FrauenLoop – organising a career panel on how to qualify for and get a job in the field.
By offering such a wide range in topics and formats, we assured that all kinds of motivations for participating in a conference would be satisfied – and it seems as if we were successful.
Don’t forget to play: Our keynote speaker Andrada brought hardware from the 80s and set up a gaming corner
“The exchange with other experts from research and industry was very valuable and inspiring”, says Hagen who enjoyed both the networking as well as the talks. “As soon as I get home I want to provision a GPU with TerraForm and try out CatBoost on it,” combining the learnings of two talks he saw (Launch Jupyter to the Cloud: an example of using Docker and Terraform, CatBoost: Fast Open-Source Gradient Boosting Library For GPU).
Talk on Terraform by Cheuk Ting Ho, one of Hagen’s favourites
“Participating in these kinds of events is indispensable for staying up-to-date with current research and state-of-the-art,” claims David who was fascinated by the interpretability offered by generalized additive models (GAMs) and is now keen on checking out the library pyGAM (pyGAM: balancing interpretability and predictive power using Generalized Additive Models in Python). Apart from feeling very happy about how much attendees and speakers enjoyed the conference and getting so much positive feedback, I was excited to get into contact with the author of spaCy, a state-of-the-art NLP library for python (Building new NLP solutions with spaCy and Prodigy), and learning on Bayesian Neural Networks, an approach I’m willing to implement for an upcoming project (Modern Approaches to Bayesian Learning with Neural Networks).
Both as committee member and as attendee, I’m very much looking forward to PyData Berlin 2019 to once again bring together so many different people in the wonderful city of Berlin.