Last week I attended the SciPy2019 conference for the first time! It was great fun, and I had an opportunity to catch up with my Elegant SciPy co-authors Juan and Stefan (only the second time we’ve all been in the same room!). These are my notes from the conference, including talks I enjoyed and tools or ideas that I found interesting.
Schedule: https://www.scipy2019.scipy.org/confschedule Proceedings (papers): http://conference.scipy.org/proceedings/scipy2019/ Talks on YouTube: https://www.youtube.com/playlist?list=PLYx7XA2nY5GcDQblpQ_M1V3PQPoLWiDAC
Day 1: July 10, 2019
Vaex: Out of Core Dataframes for Python - Maarten Breddels
Vaex - allows real-time data exploration of a DataFrame with a billion rows. Similar syntax to pandas. Creates “expressions” that are not actually executed until necessary.
To Comment or Not to Comment? A Data-Drive Look at Conflicting Attitudes Towards Commenting - Patricia Hanus
Why everyone is so afraid of commenting, and how we can use comments as a learning opportunity. Really engaging talk!
Inclusive Leadership: Engaging Contributors in the Long-Term - Tania Allard
I haven’t had a chance to watch this one yet, as I was in a concurrent session, but it got a rave on Twitter!
A Geographers Journey into AI: Mapping Urban Trees from Scratch - Verena Griess
A nice machine learning talk, well presented.
These are great fun. Many puns!
How to create a minimal python package to collect your random functions
Day 2: July 11, 2019
Generational Changes in Support for Gun Laws: A Case Study in Computational Statistics - Allen Downey
Interesting topic. Basically found that young people are less likely to support gun control. Uses linear models and talks about how to interpret them. Also comments on confounding variables.
Supporting Open Source Software for Science - J. Freeman, N. Sofroniew
Talk about the currently open Chan Zuckerberg Initiative funding call for Essential Open Source Software for Science. CZI partner with researchers to help them build open source software. Currently sppporting bioconductor, seurat and scanpy They kept funding science, or scientists, and found themselves indirectly funding open source packages. So they decided to specifically fund software!
Getting Lost in Community Building - Matthew Turk
Choose your own adventure open source project game - funny
- Slides: https://matthewturk.github.io/2019-07-11-getting-lost-in-community-building/
Inequality of Underrepresented Groups in Core Project Leadership - Anthony Scopatz
Another amusing choose your own adventure talk. I would note that the format actually works quite well to make a point about the speaker’s relationship to the topic, although the point gets a little lost in the fun a times.
Real World Numba: Creating a Skeleton Analysis Library - Juan Nunez-Iglesias
How fast can I make my code?! Note the use for actually drawing on paper in at a computer science conference!
How to Accelerate an Existing Codebase with Numba - Stanley Seibert
Similar concept to above, except taking a more pragmatic approach.
here() R function for Python - useful in Jupyter notebooks to find your data files
Day 3: July 12, 2019
Keynote: Jupyter: Always Open for Learning and Discovery - Carol Willing
Open source, open science - very thoughtful and moving talk. Book recommendation: Economics for the Common Good, Jean Tirole (or listen to the interview - link in slides) Papers: Building a Community of Open Source Contributors Ten Simple Rules for Reproducible Research in Jupyter Notebooks
Refactoring the SciPy Ecosystem for Heterogeneous Computing - Matthew Rocklin
cupy - run things on the GPU - uses the same interface to numpy and (most of) pandas.
Thoughts about how we need standards interfaces to bring together multiple backends.
Turning HPC Systems into Interactive Data Analysis Platforms - A. Banihirwe
Microscopium: Interactive Exploration of Large Imaging Datasets - Genevieve Buckley
Building and Replicating Models of Visual Search Behavior with Tensorflow, Nengo, and the Scientific Python Stack - David Nicholson
Neuro science and machine learning. Can we teach a neural network to have the same biases as a human? Cool talk!