CS seminar of Gustav Šourek
Title: Lifted Relational Neural Networks
When: [Thu, Nov 9, 2 pm]
I will talk about a framework combining deep and symbolic learning I'm developing as my PhD thesis, we call it "Lifted Relational Neural Networks" (https://arxiv.org/abs/1508.05128). It aims to marry the interpretability and expressive power of first-order logic with the effectiveness of neural network learning. In the lifted framework, first-order rules are used to describe the structure of a given problem setting. These rules are then used as a template for constructing a number of neural networks, one for each training and testing example. I'll demonstrate how this can be used for learning and reasoning from complex relational data using various modelling constructs. To spice it up I'll also share some experiences from internships at Google and IBM.