The 29th IEEE International Symposium on Industrial Electronics (ISIE), was held on the 17-19th of June 2020, as an online conference hosted by the Delft University of Technology. This was the first time this conference was held as a purely virtual event, without any physical interaction between the various participants. This proved to be an interesting and new experience, which came with its very own challenges.

Stefan Kollmann had the privilege to present a paper on Comparison of Preprocessors for Machine Learning in the Predictive Maintenance Domain by Stefan Kollmann, Alireza Estaji, Aleksey Bratukhin, Alexander Wendt and Thilo Sauter. The work covered recent findings on the performance and interactions of different data preparation- and machine-learning steps, performed as part of the ongoing research project SAVE, as was presented in the Computational Intelligence track. The paper itself was well received will be made available in the IEEE xplore online digital library and the EI Compendex database. As part of the unconventional conference format, the presentation of the paper is available in video format at loom: Loom Presentation Video

The conference itself was an interesting experience. For me, the author, this was the first fully virtual conference I attended and the additional requirements afforded by this format first became apparent sever weeks ahead of the actual event. The organisers decided on making all the paper presentations available to the participants ahead of time as screen captured videos with voice-over by the presenters. The great team of the ISIE2020 provided excellent information on how to do this and suggested very well-suited tools for creating these videos. But all their great efforts could not dampen the horror of having to record your first video-only presentation. It is surprisingly hard to explain the interactions between Preprocessing and Machine Learning to a laptop camera. But in the end, the scientist prevailed over his new technological nemesis and the presentation of our paper was even awarded Best-Presentation for the Computational Intelligence track.

The availability of all presentations ahead of conference proved a very nice side effect of the new online format and being able to browse the manifold videos via the conferences program page felt like a scientific-candy-store, where one can sample the many interesting works of their peers in their own time and speed. The presentation quality was overall good, and the participants kept surprisingly close the proposed video length of 15 minutes. During the actual event from 17-19th, sessions where held in 10 parallel Jitsi-Rooms, where presented gave another short 2 minute pitch on their paper and then answered questions from the other participants. Focus of the conference was industrial electronics and therefore my main interests, Artificial Intelligence and Machine Learning, where a little under-represented, but some of the presented works where of sufficient novelty to make the participation worthwhile.

A sharp draw back of the online format, however, where the difficulties imposed on networking and communication between participants. The question-and-answer sessions via Jitsi worked, but lacked most of the interactivity you experience in a real-world conference room. The inability to talk with a single individual, direct messaging was only available via text, and the fact that most participants de-activated their cameras, created additional distance, which made contact even harder.

So in conclusion, aside from being an opportunity to share our scientific work with community, the online conference provided interesting content in form of papers and video presentations, but fell short in terms of interacting with the researchers behind these works, which is for many scientists the main reason to attend such events.