During the 6-9th of September, Barcelona was home to the 25th International Conference on Artificial Neural Networks or ICANN, the annual conference that brings together researchers from two scientific worlds: neurosciences and information sciences. As the scope of the conference is wide, it gathers specialists in Machine Learning algorithms and computational neurosciences, as well as researchers that focus on building models of real nervous systems. This multidisciplinary conference aims to facilitate discussions and interactions in developing intelligent systems, and to enhance our understanding of the cognitive system, which makes it a great chance for NE to get involved!
This year, the conference gathered about 200-250 attendees and had a special flavor as it celebrated the 25th edition (see Figure 1 for a group picture!). This year, the conference was organized in two main tracks: Brain inspired computing and Machine learning research, where neutrally-inspired algorithms would group in neural coding, decision making and unsupervised learning approaches.
With this structure, the organizers managed to create an interactive environment with strong interdisciplinary discussions, which facilitated communication between attendees. Just to mention a few, we were able to listen Prof. Erkki Oja, one of the contributors in the design of independent component analysis (Figure 2, note that we use this algorithm daily, remember this post!); Prof. MD. Joaquin Fuster from University of California in USA with a long and intense career in the study of prefrontal cortex or Prof. Gunther Palm, an established researcher in the field of neural networks and associative memory. While there is not enough space in this blog to mention every relevant contribution to ICANN and our experience, we would like to refer you to the program of the conference where you will be able to find a list of attendees.
This time, Starlab and NE had the chance to contribute in two ways: through the organization of a special session and the presentation of our recent research. The special session gathered researchers that use advanced neural networks as pattern recognition tools for EEG/MEG, with special emphasis in deep learning and reservoir computing. Several works in the literature have been using these tools to analyze multimedia data (e.g. for speech recognition, for object detection on video), while EEG/MEG have not deserved so much attention. In our opinion, it is worth investing some effort in designing strategies tailored for EEG/MEG data, which present some special features: multi-channel temporal signals, high correlated channels, low signal-to-noise ratio, unclear feature invariants, non-linear dynamical and non-stationary signals. With this motivation in mind, we organized the special session, which was a great success!
Our special session facilitated the presentation of 6 research topics, listed in the Figure 3, which were not only scientifically relevant but also at the personal level – for the first time, I had the chance of presenting very good friends as speakers! Scientifically speaking, it was a pleasure to see how new methodologies are being used for the analysis of EEG, and I am sure that good results will soon be published and accessible for general use. You can see the presented research as conference proceedings in the Lecture Notes in Computer Science series, by Springer, as detailed in the ICANN webpage.
Did we peak your curiosity? Also, it may be a good time to start thinking about next ICANN that will take place in Alghero, Sardinia between the 11-15th of September…or to have a look at images of the first ICANN that was held in Helsinki in 1991 here.
