As a very long time member of the ACM and their SIGKDD group, I’d always wanted to attend a KDD conference (first one occurred in 1995). This year I received a gracious invitation to attend KDD2019 in Anchorage, Alaska, August 4-8. It satisfied two of my bucket list items: witnessing a KDD first-hand and also visiting Anchorage. I was not disappointed with the experience! What follows is my “Field Report,” a travel log if you will, chronicling my observations, both technical and cultural.
KDD is touted as being “the premier interdisciplinary conference bringing together researchers and practitioners from data science, data mining, knowledge discovery, large-scale data analytics, and big data.” The show was very academic, which is something you’d expect being that KDD ( vis-à-vis SIGKDD) is part of the ACM, computing’s oldest and most respected international learned society. But I learned that in the past several years the conference has developed a unique industry perspective as well. The attendees, all 3,200 of us (we increased the population of Anchorage by 1%), were evenly split between academia and industry. It made for a very well-rounded experience that I think everyone appreciated.
I started my journey from Los Angeles, my home base, and found the 5 hour flight to be nearly as long as the one to Honolulu (one of my regular holiday destinations). I’m not sure what I expected of either the conference or the venue, but I warmed up to both quickly. My hotel was a short walk to one of the two conference centers (Egan, while Dena’ina was another two blocks away). On my initial walk Monday afternoon to Egan, my stomach was empty and I was on the look out for some fast food, but I never found any of the typical American fast food establishments. Instead I came across a hot dog cart selling “Reindeer sausage” on a bun. I was in heaven.
After picking up my badge, I headed over to my first session, the “AutoML and Development Frameworks” series of talks. AutoML is a pretty hot topic, and one that resonates very well with insideAI News readers, so I was eager to hear some new perspectives. I really enjoyed the talk “Auto-Keras: An Efficient Neural Architecture Search System,” but the other 4 presentations were also quite compelling (the entire session was 2 hours). The full Proceedings for the conference including links to all accepted papers can be found HERE. In preparation for the conference, I went through the ENTIRE list of papers and selected my top 10 to print out and take along for the long flights back and forth. Here’s the list in no particular order:
- Adversarial Variational Embedding for Robust Semi-supervised Learning
- Axiomatic Interpretability for Multiclass Additive Models
- dEFEND: Explainable Fake News Detection
- ET-Lasso: A New Efficient Tuning of Lasso-type Regularization for High-Dimensional Data
- Learning Class-Conditional GANs with Active Sampling
- OBOE: Collaborative Filtering for AutoML Model Selection
- Scaling Multi-Armed Bandit Algorithms
- Tackle Balancing Constraint for Incremental Semi-Supervised Support Vector Learning
- Auto-Keras: An Efficient Neural Architecture Search System
I liked how the conference was organized around submitted research papers with live presentations by the researchers grouped by track, e.g. applied data science, research, etc. The conference also featured Hands-on Tutorials, Workshops, Deep Learning Day, Earth Day, Health Day, etc. This structure meant that each attendee had something relevant to their interests happening at all times. It was invigorating, and at some times intellectually exhausting. Maybe due to the intensity, I observed attendees in the large lecture halls engaging in short respites with laptops open doing work, checking Facebook, or in one case shopping for bathroom faucets! In the meantime, others like me were glued to the presenter’s slides, with attendees periodically raising their cell phones to take pics of slides of particular interest, making me think of a techie whack-a-mole contest.
Later Monday night (7-9pm), I attended the Poster Reception where researchers stood in front of their large academic posters explaining their research to anyone who was interested. From other conferences that had this feature, I really enjoy staring at posters for a long time, trying to soak in the main concepts. I take cell phone pics of many so I can ponder over them later. At the reception, the conference hosted a fine buffet dinner and a bar where attendees could “spend” their 4 allotted drink tickets. I sampled an Alaskan Amber which was quite comparable to my usual Sam Adams. I quickly petered out after a full day of travel so I made my way back to the hotel, in full sunlight since the sun set around 11pm.
The Location and Venue
The location for KDD2019, Anchorage, Alaska, was an interesting choice. Truth be told, this was a factor for my attendance, as I’ve always wanted to visit the city. I anticipated a chilly locale, but in reality the temps were pretty much the same as my hometown of Los Angeles (a natural desert). The effects of climate change are clear in our 49th state with July being their hottest month on record, and Anchorage experiencing its hottest day ever at 95 degrees. One surprise was the 16.5 hours of sunlight each day. It gave me and other KDDers plenty of time to play and explore once the conference day was over.
I found out that many conference goers discovered the 49th State Brewing Company, only a block away from the Hilton Anchorage where I was staying. It must be a good hang-out, because early Wednesday morning I woke from my sleep at 2:30am and glanced out my window only to see a lone female KDDer (with her badge dangling around her neck) trudging away from the bar back to the hotel. Now that’s the way to enjoy a tech conference!
I hear that KDD2020 will be much closer to me next year, in sunny San Diego, Calif., although the 125 mile drive from LA once took me LONGER than the flight to Anchorage!
The Exhibition Hall
The Exhibition Hall (which doubled as the lunch venue) was well-represented with industry vendors like Microsoft, IBM Research AI, Yahoo Research, Neo4j, TigerGraph, KenSci, Gurobi, HexagonML, and many others. I love kicking around conference exhibition halls because I always learn about companies not currently in the massive insideAI News industry database. At KDD2019 I learned about two hardware companies: Lambda Labs (I’m now wearing their cool t-shirt to the gym) with their deep learning machines, and also Inspur AI, both providing systems for AI workloads.
Surprisingly for a conference like KDD, I also saw a number of companies that “use” data science and machine learning such as Pinterest, Lyft, Etsy, Snap, Capital One, Home Depot and Target. These companies were there presumably to recruit new members for their data science teams, and there were plenty of quality candidates present. I think it was a good move for such companies to show off how leading-edge they are with respect to all that’s data.
I was pleased that book vendors showed up in force: MIT Press, Cambridge University Press, Springer Nature, and CRC Press. Being in one place, I was able to review many new titles for potential book reviews, as well as my own professional use. I was told by one publisher that they don’t like to schlep the heavy books back home so they try to leave them with local universities, but there were no takers in Anchorage. I wonder if that had anything to do with the enormous funding cuts the University of Alaska is grappling with?
Chance Encounters
Wearing my journalist hat, I typically run into many fascinating people at industry conferences. At lunch on Tuesday, I was able to kibitz with a number of attendees at my table including some who came from great distances. One data scientist I met came from Brazil, involving a 25 hour trek. That’s commitment!
At lunch on Wednesday, I had a nice chat with “David” from Pinterest. He intimated how his company blurs the lines between data scientists and data engineers, and that he frequently has to switch-hit on projects. I would have thought a newly public company would have worked to have more defined roles, but I understand how the unicorn mystique is hard to break, especially with high-flying tech companies.
Tuesday morning, I had a nice breakfast in the lobby restaurant at the Hilton. I was quite pleased to find “reindeer sausage” on the menu! My waitress was a 20-something woman with an unfamiliar accent. I found out she was only a month in Anchorage, part of a cultural exchange program from Serbia. I asked what she was studying at her university, and was delighted to hear “computer science.” She was unaware that KDD was in town, and became excited to learn that several thousand computer scientists were in town.
At the second Poster Reception on Tuesday evening, I ran into a friend, Jon Morra, who is the organizer for the LA Machine Learning meetup group. We chatted about all the cool research being presented at the show, and he mentioned BERT as being quite hot. I enjoy chance meeting like this to point me in new directions. Back in my hotel room that night I downloaded and started reading a paper about BERT. Check out this seminal paper “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” from Google AI Language which started the ball rolling.
The Road Home
On my flight back to LA on Wednesday evening, I spent the long trip going through a recent MIT Press title “Introduction to Deep Learning,” by Eugene Charniak that I’m reviewing for insideAI News which includes a lot of math. It was a great way to wind down and relax after an exhilarating few days at KDD where I learned so much.
The very next day upon returning to my data science practice, I ordered a case of Alaska Glacier water (source is the Eklutna Glacier located high in Alaska’s Chugach Mountains). I bought a bottle at the Anchorage airport just before my flight home, and it was delicious. Glaciers are becoming a thing of the past due to anthropogenic climate change so I thought I’d load up and use it as a reminder of my great time at KDD.
I would highly recommend this event to anyone in the field, or desiring to enter it. I believe I’ll have to add KDD to my list of annual conferences for sure!
Contributed by Daniel D. Gutierrez, Managing Editor and Resident Data Scientist for insideAI News. In addition to being a tech journalist, Daniel also is a consultant in data scientist, author, educator and sits on a number of advisory boards for various start-up companies.
Sign up for the free insideAI News newsletter.
Speak Your Mind