Join the conversation and learn more with José Amorim, from the University of Coimbra, and with Jafar Adibi, Talkdesk's Head of Data Science and AI.
Talk #1
Title: Interpreting Deep Learning
Description: While deep learning models have reached high predictive capabilities in a wide range of applications, they are often perceived as black-boxes. Recently a variety of methods for interpreting these models have been introduced. While feature importance and visualizations give insight into the model’s decisions, rule extraction methods help us understand the model’s logic. Nevertheless, the tradeoff between performance and interpretability is still an open problem. This talk will start with a brief overview of deep learning, followed by a discussion of state-of-the-art methods used to interpret deep learning models and decisions.
By José Amorim (University of Coimbra)
Bio: José Pereira Amorim is a PhD student at the University of Coimbra. He got his master’s degree in Informatics Engineering at Faculty of Engineering of the University of Porto (FEUP). José is currently affiliated with the IPO-Porto Research Centre (CI-IPOP) and the Centre for Informatics and Systems of the University of Coimbra (CISUC) and his research is mainly focused on creating interpretable models in order to assist physicians in the field of oncology.
