The unconference1 was unlike any event I have been to in the past. 60 participants from around the world got together, formed small groups based on their common interest in a github issue, and worked on a project for 2 days.
I joined up with the UMAP group to figure out a way to implement the new dimensional reduction algorithm UMAP in R (similar to t-SNE, but potentially faster!). We ended up wrapping the Python implementation of UMAP using the
reticulate package, which was actually quite easy! The next step is re-writing the entire algorithm in Rcpp, but …….. luckily it looks like someone else has already been working on that!
I was intimidated at first by the caliber of all the other amazing participants, but the organizers were really good at making everyone feel welcome and as prepared as possible. It was clear that they intentionally chose to go with a small format in order to foster a community of trust between like-minded people. If I could do it again, I would not have worried about knowing less than everybody else or getting in the way of others because most people were really excited to share knowledge about R and help each other out.
1unconference. I wasn’t sure if it was a real word…
Cascadia R Conf
I also attended the second annual Cascadia R Conf this past weekend. I was impressed with how much they were able to pack into a one-day conference.
Some of the highlights for me were:
- Kara Woo’s Keynote, which was a great introduction to open source/github contribution culture. Plus I learned about the reprex package for creating minimal reproducible examples in markdown format.
- Jonathan Nolis’s Using deep learning and R to generate offensive license plates was pretty funny as well as informative. The Machine Learning workshop was also good for making TensorFlow seem a little more tangible to me (I still have a long ways to go, but apparently TensorFlow is linear regression and linear algebra at its core).
- The cRaggy graphics show-and-tell was an interesting format in which participants downloaded a dataset (Portland’s Biketown bike share system usage data), visualized it using any method they like, and shared their results on the conference day. I enjoyed Charlotte Wickham’s example, which involved clustering bike hubs by time of trip and plotting the clusters using ggmap. I’m gonna check out the similar format Tidy Tuesday for future projects.
There was also a panel discussion about imposter syndrome and “gatekeeping”, which is “The activity of controlling, and usually limiting, general access to something”2.
One of the points which I found interesting was the acknowledgement that imposter syndrome and gatekeeping can occur at the same time, meaning that both internal and external factors contribute to feelings of inadequacy.
During the question and answer session I wanted to ask “How does someone differentiate between imposter syndrome and incompetence?”, but I thought it might be an awkward question.
It looks like I’m not the first person to ask this question, though. I personally think it’s a bit of a grey scale.
It was valuable to meet new people, learn about lots of packages, and see what topics other people in the R community are interested in. I’m thankful to the (un)conference organizers for giving me the opportunity to attend the events.