177 Huntington Avenue
Boston, MA 02115
Heiko Zimmermann is a PhD student at Northeastern’s College of Computer and Information Science focusing on machine learning, advised by Professor Dr. Jan-Willem van de Meent. He works on probabilistic modeling and Bayesian inference. He comes from Stuttgart, Germany.
Zimmermann earned his BS and MS from the University of Stuttgart, where he worked on Bayesian Optimization and its applications in Robotics and biomechanics. At Northeastern he is focusing on Bayesian methods for inference in probabilistic models, probabilistic programming, and efficient inference techniques.
- BSc, University of Stuttgart
- MSc, University of Stuttgart
- Hometown: Stuttgart, Germany
- Field of Study: CS, Machine Learning
- PhD Advisor: Jan-Willem van de Meent
What are the specifics of your graduate education (thus far)?
My graduate research at the Machine Learning and Robotics Lab in Stuttgart focused on Bayesian Optimization including theoretical aspects, extension of Bayesian Optimization to functional domains, and applications in robotics and biomechanics. At Northeastern, I’ll be working on probabilistic modeling and Bayesian inference.
What are your research interests in a bit more detail? Is your current academic/research path what you always had in mind for yourself, or has it evolved somewhat? If so, how/why?
Our world involves uncertainties on many different levels. When modeling different aspects of the world, we need to capture these uncertainties to answer questions and quantify our confidence in terms of probabilities. Probabilistic modeling and inference are at the core of this.
What’s one problem you’d like to solve with your research/work?
In Machine Learning and AI, a lot of time is spent on finding the right model and inference/optimization technique for specific tasks. This involves the tuning of hyperparameters but also fundamental decisions regarding the model’s structure that are tightly coupled to the availability and choice of inference algorithms. Probabilistic Programming is a framework that enables domain experts to build models in a well understood way, in the form of a program, without in-depth knowledge about the inference. I would like to contribute to the ongoing effort to further automate probabilistic modeling and inference.
What aspect of what you do is most interesting/fascinating to you? What aspects of your research (findings, angles, problems you’re solving) might surprise others?
The principles that underpin Bayesian inference are remarkably simple, however, for a lot of interesting setting the simple application of these principles results in intractable problems. Working around these intractabilities is often a tricky task.
What are your research/career goals, going forward?
Establishing myself as a researcher.