user! 2019: What to Expect and How to Prepare

For the first time this year, I will participate to the useR! conference.

I’m not exactly sure what to expect.

Last time I went to a similar conference, it was the SSC (Statistical Society of Canada) annual meeting of 2015.

Back then, I was still a PhD student.

I didn’t understand most talks I went to.

And, to be honest, a lot of them weren’t attractive to me.

But this time, it’s different.

This time, I chose to go to useR!2019 because I’ve been using R for many years and that’s the technology I have used with all my past employers and clients.

Now, I feel I’m part of this world.

There are a few things I expect from this conference, and I would like to prepare some goals as well.

I want to:

• learn
• discover
• meet

Let’s dive into each goal.

1. Tutorials: Learn

This list of tutorials is here: Tutorials

Pretty much every day, I encounter something I don’t know in R.

There are things I know very well, I have become an expert at them: dplyr, data.table, shiny, …

There are things I know and I may even have used them, but I definitely don’t master them: plumber, golem, shinyproxy, …

And finally, there are things I know nothing about: H20, bookdown, purrr, …

It’s hard to know everything.

But it’s also hard to deal with this imposter syndrome because I know the huge amount of things I don’t know.

And finally, it’s also hard to take the time to sit down, open a book (or a tutorial, or even a package vignette), and LEARN.

I expect to learn during useR:2019.

I subscribed to two tutorials:

Automatic and Explainable Machine Learning with H20 in R.

I know quite a lot about machine learning.

But I know nothing about H20.

Actually, I don’t know much about AutoML in general.

I feel this is wrong.

So I hope the 4 hours tutorial on H20 will at least help me get a good grasp on the strengths of AutoML and teach me how to apply it for my clients.

Docker for Data Science: R, ShinyProxy and more.

I know some of Docker.

Mostly pulling images and setting them up on a server.

But I have struggled to setup ShinyProxy in the past and make it work perfectly.

Even though..

I created many Shiny apps for my clients and hosted them on a server (without Docker usually).

Again, I feel I would be so much efficient if only I could learn this.

And rather than spending days figuring everything out by myself, I better learn it from an expert.

This first day of tutorials will already teach me things I have been wanting to learn for MONTHS (if not years).

2. Talks: Discover

The list of talks with the abstracts: Talks

Talks will take place from Wednesday to Friday.

Reading through the talks subjects and descriptions reminded me of my time as a PhD student.

I would look at all the talks from the conference, trying to decrypt the titles and predict whether I would understand something after the first 3 slides.

It’s so different now!

Looking at all the talks, I felt EXCITEMENT!

I felt I wanted to go to all of them!

Ok, ok, maybe not all of them.

But at almost every timeslot, I could find something exciting to look forward to.

Wednesday Morning: Shiny 1

Shiny 1 is about:

• Logging and Analyzing Events in Complex Shiny Apps
• mwshiny: Connecting Shiny Across Multiple Windows
• Shiny app deployment and integration into a custom website gallery
• Automated Surveys and Reports for Expert Elicitation with Shiny

I know it’s possible to switch between rooms to change the theme, but from experience, it’s a bit hard to do without missing a talk. I prefer to stay in the same room for the duration of the batch of talks.

In this first batch, it’s all about Shiny.

That just makes sense since many of my clients want something with Shiny (who wouldn’t?).

I am especially curious about the last one: Automated Surveys and Reports for Expert Elicitation with Shiny. Looking at the abstract, this is super close to some work I have done with a past client. I am very very curious at what they have done.

It’s also funny that both (my client and the speaker) chose Shiny for this, as it’s not especially R related. There is not a huge amount of moving data around. A little bit of statistics, okay, but not that much. It’s a web app. But R Shiny is actually very strong at just building a web app.

Wednesday Afternoon: Shiny 2

Shiny 2 is about:

• Golem: A framework for Building Robust & Production Ready Shiny Apps
• Art of the Feature Toggle: Patterns for maintaining and improving Shiny applications over time
• Data for all: Empowering teams with scalable Shiny applications
• Best practices for building Shiny enterprise applications

Yup.

Shiny all over again.

I’m mostly curious about the Golem talk.

I only started to use it recently, after having glanced over it for months.

It felt both practical and clumsy. Probably because I need to get used to it. There are many functions in golem and the accompanying packages I don’t know, so that’s probably why I need to get used to it. Also, as they state, it is an opinionated framework.

Getting re-introduced to it by an expert will certainly help to acquire the good practices more quickly.

Thursday Morning: Text mining

Text mining is about:

• {polite} - web etiquette for R users
• The R Package sentometrics to Compute, Aggregate and Predict with Textual Sentiment
• BibliographeR: a set of tools to help your bibliographic research
• ggwordcloud: a word cloud geometry for ggplot2
• Die nutella oder Das Nutella? Grammatical Gender Prediction of German Nouns
• Implementing a Classification and Filtering App for Multilingual Facebook Comments - A Use Case of Data for Good with R

I started in text mining with a client who is a researcher in political science.

In the beginning, I was mostly doing web scraping, and more and more I have learned and implemented text mining algorithms.

In this batch of talks, I am mostly interested in polite and the one about German nouns. The former because I realize when I scrape that.. I just do it brutally. I don’t have much concern for the servers I’m scraping, and I realize there must be a better way.

The latter because I know some German and I know how hard it is to remember the genders of all the nouns. I also admire the creativity in thinking one might try to predict the gender of a German name. As a human being, I have asked myself whether there was some logic that a noun should be masculine, feminine, or neutral, and never found one. It’s not even coherent between French and German. So.. I’m curious here.

Apparently, these are very short talks, so it’s mostly about hearing about something, taking some notes, and diving into it later at home if it spikes my interest.

Thursday Morning later: Data mining

Data mining is about:

• Machine Learning with R: do it with a framework
• Building and Benchmarking Automatic Machine Learning Systems
• mlr3: A new modular framework for machine learning with R
• mlr3pipelines: Machine Learning Pipelines as Graphs

Machine Learning is also a topic of interest of mine, and something I sell to my clients. So far, I have either done it with Python or with the caret package. Or, actually, with the so many different packages all specialized on ONE type of algorithm.

It never felt as easy as with scikit-learn in Python.

So I want to learn more about what exists.

During this session, there is also a talk about R and security. I would really like to listen to this one, as they will talk about how to hack into R. I am concerned about security whenever I host an R API or an R Shiny but don’t know much about it.

Thursday Afternoon: Operations & data products

Operations & data products is about:

• How a non-profit uses R for its daily operations
• rjenkins and rrundeck: Coordinating Continuous Integration and Delivery with R
• Advanced Git Integrations for Automating the Delivery of Reproducible Data Products in R
• Github actions for R

These are not the most exciting talks I want to listen to, especially since I’ll be quite exhausted after 4 days of conferencing, but I see it as some nice bonus.

Continuous Integration and automating tasks is something I know would make my work more efficient.

3. People: Meet

Finally.

One of the best perks of going to a conference is meeting people.

So far, I don’t know I know anyone going there, so that’s a bit scary for me.

I don’t go to people that easily.

But I still hope to meet people and learning about the R environment in France.

In fact, I don’t even know the kind of people I’ll find there.

Mostly academics?

Probably some, but I believe a lot of companies now use R. I can just look at the sponsors.

One team I really expect to meet there is ThinkR.

They’re 5 R experts who founded an R Agency a few years ago and they seem to be growing a lot. I see them everywhere, they develop super useful tools (like golem), so I’m curious to meet and learn from them.

Another team I’d like to meet is Ardata.

I know a few other data scientist freelancers as well (such as Dacta) that will go there and that I would like to meet.

It’s hard to set a goal for meeting people, but I expect to learn a lot about the R ecosystem and who are the R users.

More about this when I come back (;