A post from July 2020, lightly copy-edited and preserved here from the original blog.
What SciPy 2020 was
SciPy is a community dedicated to the advancement of scientific computing through open-source Python software for mathematics, science, and engineering.
The annual SciPy Conferences let participants from academic, commercial, and governmental organizations:
- Showcase their latest scientific Python projects.
- Learn from skilled users and developers.
- Collaborate on code development.
The conferences usually consist of a few days of tutorials, followed by two or three days of presentations, and conclude with one or two days of developer sprints on projects of interest to the attendees.
This year the format was different — the structure was completely restructured to adapt to the global situation of the COVID-19 pandemic. Even with this big change (and my first experience with the event), I was amazed by the engagement of the community: a rich program, reviewed submissions, recruited tutorials, and managed sprints, all with very little lead time.
As the conference’s Welcome Message put it:
They figured out how to retain much of the atmosphere that we appreciate about SciPy meetings, while accommodating the constraints and possibilities of a virtual conference.
Tutorials
The conference hosted a wide range of subjects for all levels of programming experience. There was a clear effort to gradually involve the participant — so even if you were a beginner Python programmer (or an advanced one — there’s always something new to learn) you could follow the subjects and experiences as they came.
It started with a setup tutorial covering the must-have tools — Python, conda, and the most-used packages. There were also introductory sections on Python and programming, which I thought was really nice: in my experience with mathematics conferences, attendees usually self-select for subjects they already understand. An introductory level inside an international conference is an incredible accessibility move.
Then the welcome presentations and plenary sections introducing the scheduled tutorials.
I couldn’t watch all the presentations because of some personal matters, but a few of the most interesting tutorials I did catch:
- Machine Learning Q&A and Moderated Discussion (didn’t finish)
- Bayesian Data Science by Simulation Tutorial (as a mathematician, couldn’t miss this)
- Parallel and Distributed Computing in Python with Dask Tutorial (fantastic)
- Introduction to Conda for (Data) Scientists Tutorial
- Biology and Bioinformatics (talks and discussions)
- Deep Learning from Scratch with PyTorch (still need to watch)
Two stood out for teaching quality and materials: the Bayesian Data Science by Simulation tutorial by Eric and Hugo — excellent presentation — and Parallel and Distributed Computing in Python with Dask, which came with an extremely well-prepared Jupyter notebook that the presenters used to follow their comments live, and they answered nearly every question from the community in real time.
A great number of tutorials, and an even bigger effort to help everyone with their projects and shared experiences.
Birds-of-a-Feather sessions
To increase community-building opportunities, SciPy emphasizes Birds of a Feather (BoF) sessions. They usually include short presentations by a panel and a moderator, with most of the time spent opening up discussion to everyone in attendance. BoF topics can be general (state-of-the-project) or based on the themes of the conference and the mini-symposia.
This year’s BoF topics:
- Python in Education for Generation Z
- Packaging
- Diversity and Inclusion: Incident Response
- Driving Hardware for Instrumentation
- SciPy 2021 Planning
I could only watch Packaging, Driving Hardware for Instrumentation, and SciPy 2021 Planning. The standout for me was Birds of a Feather: Packaging — it covered Anaconda, conda-forge, and information about the current and future structure of conda-forge.
My mentors C. J. Wright and Marius van Niekerk presented information about conda-forge and answered questions from the community. There were also points from the Anaconda team and other institutions that host packages on conda-forge. The full discussion is on YouTube: Birds of a Feather: Packaging.
Conclusion
As my first international event — and my first time interacting with so many different communities (Anaconda, SciPy, NumPy, conda-forge) at once — the reality went well beyond my expectations. The structure of the event itself was a small marvel: the ability to change between “rooms” for different presentations the way you do in physical attendance, plus the integration of Jupyter notebooks, terminals, and other tools that helped familiarize the viewer (and the student — in my case) with what was being demonstrated.
It created an environment that, at moments, felt like being back at university — surrounded by friends, with a vast number of subjects to choose from and dive into.
The current structure of the conference opens a big door of possibilities, not just for fast talks but for a genuinely new way for programmers to show their projects and research to a global community.
The conclusion of my experience: it’s an excellent way to meet new projects and applications worldwide, and a real stepping stone for anyone wanting to learn something new from the hands and minds of the people who built it. I am very thankful to the conda-forge and NumFOCUS team for the chance to participate. I really hope to go to SciPy 2021 — ideally in person — and meet everyone properly, to talk and discuss the future ideas and questions in the room together.
For more on the SciPy conventions and to watch some of the tutorials and discussions from SciPy 2020, check the EnthoughtMedia channel on YouTube.