Phillip Swazinna

Reinforcement Learning Research Scientist

Siemens

Technical University of Munich

Biography

Ever since I went to UPenn as an exchange student in 2015, I have been fascinated by different AI topics. I wrote my Bachelor’s Thesis about statistically modeling attention on scientific publications and my Master’s Thesis on using CNNs to classify various neurodegenerative diseases in MRI scans with weak supervision. In my PhD, I have shifted focus again, and developed multiple algorithms for model-based offline reinforcement learning, which is a topic I remain passionate about.

Interests

Offline RL
Reinforcement Learning
World Models

Education

Ph.D. on Model-Based Offline RL, 2023

Siemens & TU Munich
M.Sc. Informatics - AI & Robotics, 2018

TU Munich & ParisTech
B.Sc. Informatics, 2016

TU Dortmund & UPenn

Recent Publications

Negative-Unlabeled Learning for Diffusion MRI

Phillip Swazinna, Vladimir Golkov, Ilona Lipp, Eleonora Sgarlata, Valentina Tomassini, Derek K. Jones, Daniel Cremers

PDF Poster

q-Space Deep Learning for Alzheimer’s Disease Diagnosis: Global Prediction and Weakly-Supervised Localization

Vladimir Golkov, Phillip Swazinna, Marcel Schmitt, Qadeer A. Khan, Chantal M.W. Tax, Marat Serahlazau, Francesco Pasa, Franz Pfeiffer, Geert Jan Biessels, Alexander Leemans, Daniel Cremers

PDF Poster

Experience

April 2019 – Present

Munich

PhD Candidate in Deep Multi Task Reinforcement Learning

Siemens & Technical University of Munich

My research interests include:

Multi Task Reinforcement Learning
Intrinsically motivated agents
Hierarchical Methods

June 2018 – November 2018

Munich

Working Student

Siemens - Learning Systems Group

Responsibilities included:

Execution & evaluation of Deep Learning experiments with Python, TensorFlow & Keras
Implementation & testing of hierarchical classifiers based on Convolutional Neural Networks for Medical Image Classification
Test of new Bottleneck methods to reduce the amount of parameters needed in models
Evaluation of recently published research results in the area of Deep Anomaly Detection

February 2017 – April 2017

Dortmund

Internship

Materna Information & Communications SE

Development of ChatBot Applications with

IBM Bluemix
Microsoft Azure
Javascript, node.js

September 2014 – August 2016

Dortmund

Teaching Assistant

Technical University of Dortmund

Teaching Assistant for the Undergraduate Classes:

Computer Structures
Data Structures, Algorithms & Programming 2
Information Systems