In this blog post, I will talk about the IEEE ICDM 2020 conference that I have attended virtually. The conference was supposed to be held in Italy but due to the coronavirus pandemic, it was held online.
About the ICDM conference
This year was the 20th edition of the IEEE ICDM conference. It is a well-known conference that is quite competitive. It is one of the top data mining conferences. The proceedings are published by IEEE. The conference has a research paper track, as well as a dozen workshops and tutorials. The papers are mainly about data mining and machine learning.
The first day was mainly for workshops. On the second day, there was the ICDM conference opening. In the opening, the organizers were introduced, and an overview of the conference was given. Here are some of the slides, below.
The main research topics this year were:
Some statistics about the review process and accepted papers:
Most accepted papers are from China and the US, followed by Australia, Germany, India, France and Japan.
The online conference system
The ICDM conference was held on the website Underline.io where the prerecorded videos of papers could be viewed at anytime. Then, during sessions of the conferences, authors would join a Zoom session and give a 3 minutes summary of their papers and answer questions live, assuming that people had watched the videos already. A few sessions like the conference opening ceremony were held live.
Besides, there is an interesting function on Underline called the Lounge, implemented in with Gather.town, which allows to perform a video\audio chat with other conference attendees in a game-like virtual world (see picture below). In the lounge, the chat function is proximity-based. You can move your avatar close to the avatar of other persons to initiate a discussion with that person or listen to a discussion.
This is an interesting concept that aims to recreate how people would talk with each other during the coffee breaks of an on-site conference. However, in practice, there was not so many people in the lounge. I checked a few times during the first days of the conference and there was about 3 to 5 persons there。 But no one was discussing with each other. So it seems that this function is an interesting concept but in practice I did not see it being used.
My opinion about Underline is that it is relatively simple and it did the job but it relies on external services such as Zoom and Gather.town. Thus, Underline is more like a hub for different services for the conference. Having all these services under a single website or software would have been better in my opinion.
The registration of ICDM was quite low this year at 500 $ USD due to the conference being held online (because of the coronavirus pandemic). This is appreciated as ICDM is typically quite expensive, just like some other top conferences.
UDML 2020 worksop on Utility Pattern Mining and Learning
This year was the third edition of the UDML workshop on Utility Driven Mining and Learning (UDML 2020). This year, eleven papers were submitted and five were selected for publication for an acceptance rate of 45%. Three of the selected papers are about algorithm for high utility pattern mining, while another is related to spatiotemporal data mining, and another about multi-objective recommendation.
Here is a picture of the five accepted papers:
There was a good discussion during the workshop and it was nice to see some researchers that I knew already.
If you want to see the video of the paper of my post-doc about mining cross-level high utility itemsets, you may watch the video here.
Retrospect about 20 years of ICDM
To close the ICDM 2020 conference, there was a panel about ““20 Years of IEEE ICDM: Retrospect and Prospect” to discuss the two decades during which that conference was held.
Next year: ICDM 2021
The conference ICDM 2021 will be held in New Zealand, organized by the University of Auckland, from December 7 to 10 2021.
There was numerous paper presentations on various topics. I have listened to a few of them related to my interests. On some sessions, there was several watchers and several questions were asked.
Philippe Fournier-Viger is a professor of Computer Science and also the founder of the open-source data mining software SPMF, offering more than 170data mining algorithms.