Learning with logic and neural networks - Gusta Šourek
Recently, neural networks have witnessed a remarkable resurgence of interest, yet the current architectures, no matter how deep, still possess many principled limitations. One of them is the limited ability to work with structured input data, for instance, if the inputs have the form of graphs or if they reside in a relational database. Examples of such data are abundant, including databases of organic molecules, social networks, engineering designs etc. Likewise, if symbolic background knowledge is available, e.g. in the form of first-order logic (FOL) theories, there is generally no principled way to use it for training.I will talk about a logical approach to machine learning and more particularly "Lifted Relational Neural Networks" combining FOL and neural nets in a principled way, allowing to train deep networks from arbitrarily structured data, while being able to merge with and exploit symbolic expert knowledge.
Speaker: Gusta Šourek (https://www.linkedin.com/in/gustav-sourek-284a5716/ (https://www.linkedin.com/in/petr-baudis-906a213/))
• 18:00 The talk
• 20:00 - ... Networking
Machine Learning Meetups (MLMU) is an independent platform for people interested in Machine Learning, Information Retrieval, Natural Language Processing, Computer Vision, Pattern Recognition, Data Journalism, Artificial Intelligence, Agent Systems and all the related topics. MLMU is a regular community meeting usually consisting of a talk, a discussion and a subsequent networking allowing people to network, inspire each other and learn about exciting stuff. At the end of the year 2016, MLMU spread also to Brno and Bratislava. The beginning of 4th season brought MLMU Košice!