Rough Cognitive Networks
You are cordially invited to the upcoming technical talk:
Speaker: Gonzalo Nápoles, Faculty of Business Economics, Hasselt University, Belgium
Co-organized by: IEEE CI/SMC Ottawa Joint Chapter (http://www.ieeeottawa.ca/ci) and University of Ottawa Computer Science Graduate Student Association
Where: School of Electrical Engineering and Computer Science, Room SITE 5084, 800 King Edward Ave, Ottawa
When: Wednesday August 8th, 2018, 6:30 PM– 8:30 PM
6:30 – 6:45 PM Pizza / drinks and networking
6:45 – 8:15 PM Technical talk
8:15 – 8:30 PM Q&A / networking
Admission: Free but registration is required via Eventbrite (https://www.eventbrite.ca/e/technical-talk-rough-cognitive-networks-tickets-48244520660)
The ease with which we recognize our beloved black cat from hundreds similar to it or read handwritten characters belies the astoundingly complex processes that underlie these common scenarios. Which is why researchers in Machine Learning have been focused on developing a wide range of classification algorithms called classifiers with the goal of tackling these situations with the best possible accuracy. However, most accurate classification models regularly perform like black-boxes, thus neglecting the premise that understanding is an essential part of any learning process, even when the process itself is quite subjective.
In this talk, we will discuss how to exploit information granules in the form of rough sets to design a transparent neural classifier. This model comprises three well-defined steps that are materialized through the Rough Cognitive Networks. In the first step, we discover information granules on the available information using the Rough Set Theory. In the second step, we build a Fuzzy Cognitive Map where input neurons represent the previously discovered granules, while output ones denote the decision classes to be considered. The last step focuses on performing the neural reasoning process using intelligible inclusion equations and causal relations.
Rough Cognitive Networks are capable of computing high-quality predictions (when compared to traditional classifiers) even using a very small network topology. More importantly, the weights are prescriptively established, no further learning is required.
Gonzalo Nápoles received his PhD degree from Hasselt University (Belgium) and Maastricht University (The Netherlands) in 2017. He has published several papers in peer-reviewed journals including IEEE Transactions on Fuzzy Systems, Neurocomputing, Neural Processing Letters, Information Sciences, Neural Networks, Knowledge-based Systems, among others. Dr. Nápoles was recipient of the Cuban Academy of Science Award twice (2013 and 2014), the highest academic award in Cuba. He is the senior developer of the FCM Expert software tool (www.fcmexpert.net) for fuzzy cognitive maps. His research interests include cognitive mapping, rough cognitive networks, neural networks, learning systems and chaos theory.