My name is Martin Trapp, I’m a PhD student and researcher working in machine learning. I have a particular interested in tractable deep probabilistic models such as sum-product networks, on Bayesian nonparametric approaches, probabilistic programming and Bayesian learning. My thesis is supervised by Prof. Franz Pernkopf from the Graz University of Technology and in cooperation with Dr Robert Peharz from the University of Cambridge. Until Mai 2019 I was also affiliated to Austrian Research Institute for Artificial Intelligence led by Prof. Robert Trappl. I try to make all my research open-source. You can find all the packages I have developed during the last years on my GitHub account.
Additionally, I’m a member of the Turing.jl team which aims to provide an efficient, robust and intuitive probabilistic programming language written in Julia.
- [June 2019] I presented our work Turing.jl for Probabilistic Programming with Discrete Random Measures with Emile Mathiu, Maria Lomeli and Hong Ge at the International Conference on Bayesian Nonparametrics
- [May 2019] Our paper on Random Sum-Product Networks: A Simple and Effective Approach to Probabilistic Deep Learning with Robert Peharz, Antonio Vergari, Karl Stelzner, Alejandro Molina, Xiaoting Shao, Kristian Kersting, and Zoubin Ghahramani has been accepted at UAI (acceptance rate 26.2%)
- [May 2019] Our paper on Optimisation of Overparametrized Sum-Product Networks with Robert Peharz and Franz Pernkopf has been accepted at the 3rd Workshop on Tractable Probabilistic Models at ICML
- [May 2019] Our work on Bayesian Learning of Sum-Product Networks is available as pre-print
- [June 2018] Our work on Learning Deep Mixtures of Gaussian Process Experts using Sum-Product Networks with Robert Peharz, Franz Pernkopf and Carl E. Rasmussen has been accepted at the 2nd Workshop on Tractable Probabilistic Models at ICML
- [June 2017] I presented our work on Infinite Sum-Product Trees at the International Conference on Bayesian Nonparametrics
- [June 2017] Our paper on Safe Semi-Supervised Learning of Sum-Product Networks with Tamas Madl, Robert Peharz, Franz Pernkopf, and Robert Trappl has been accepted at UAI (acceptance rate 31.0%)