Workshop – Probabilistic Programming for Machine Learning

Workshop Probalistic programing

Workshop Probalistic programing

Probabilistic Programming for Machine Learning

 Recently, the plethora of deep learning methods have achieved unequivocal success on various tasks, however all of them glance over a key point in order to be reliably deployed on autonomous agents – the environment of an agent is uncertain. This uncertainty is caused by stochastic processes, sensory noise and motor noise. Probabilistic models offer a rich toolbox for representing a vast range of problems, embracing uncertainty in a rigorous and interpretable manner. However, designing and implementing a scalable inference procedure is often difficult, error prone and sometimes even intractable.  Probabilistic programming attempts to resolve those issues by merging the gap between machine learning, probability theory and computer science. Probabilistic programming languages providing flexible programmable inference engines which are capable of performing both exact and approximate inference. In this workshop you will learn how to cast any machine learning problem into a probabilistic model. We will have a look at the probabilistic language Anglican and implement several probabilistic models ranging from coin flipping to intuitive physics.,

About Svetlin Penkov

Robotics has inspired Svetlin imagination ever since he was a kid. In his journey towards creating intelligent robots he has studied Robotics at the University of Reading, UK, and Neuroinformatics and Computational Neuroscience at the University of Edinburgh, UK, where he is currently in the final stage of his PhD on Robotics and Artificial Intelligence. His research is focused at the intersection of machine learning and robotics attempting to enable robots to function in complex environments that require close interaction with people. Currently, he also spend large part of his time as a research scientist at the start up company FiveAI, where they aspire to build safe and robust autonomous vehicles.

Installation instructions

 During the workshop we will be using Anglican which is a probabilistic language written in Clojure. Therefore, you should install:

  • Latest Java JDK and Java JRE on you machine
  • Leiningen, which is a build tool for Clojure. Simply follow the instructions here:
  • Files repository GitHib repo
  • Please confirm your attendance, there are limited places
  • The location is at Sofia University, Faculty of Economics and Business Administration
    1113 Sofia, 125 Tsarigradsko Shosse Blvd., bl.3
  • Hall 200


– 14:30 – 15:00 – laptop setup

– 15:00 – 18:00 – workshop

– 18:00 – 20:00 – networking and beer, sponsored by the laziest

Share this

Leave a Reply