Fabian Falck

Fabian Falck

PhD student in
Statistical Machine Learning

University of Oxford, OxCSML

Hi and welcome! I am Fabian - great to have you here!

I am a final-year year PhD student in Statistical Machine Learning in the OxCSML group at University of Oxford, supervised by Prof. Chris Holmes and Prof. Arnaud Doucet. I am also a Graduate Teaching and Research Scholar in Computer Science at Oriel College, University of Oxford, teaching maths courses.

My research interest lies in generative modelling, and intersects with probabilistic deep learning, causal inference, Bayesian inference, unsupervised representation learning and its applications in health.

Recently, my research analysed why U-Nets are a useful inductive bias in diffusion models (NeurIPS 2023), connections between hierarchical variational autoencoders and diffusion processes, as well as U-Nets and Wavelets (NeurIPS 2022 oral), and variational autoencoders with a structured prior for clustering to find multiple partitions of high-dimensional data (NeurIPS 2021). I also studied generalising the propensity score theory to balancing scores in matching for treatment effect estimation in causal inference (AISTATS 2022). In the summers of 2022 and 2023, I interned with the Amazon Web Services AI Lab in Berlin, and Microsoft Research in Cambridge.

Before joining Oxford for my PhD, I worked with Prof. Andrew Davison in computer vision and robotics at the Dyson Robotics Lab, and with Dr. Petar Kormushev at the Robot Intelligence Lab, both at Imperial College London. I also worked with Prof. Artur Dubrawski on machine learning for health at the Auton Lab at Carnegie Mellon University. Selected publications include a comparison of view-based and map-based semantic labelling in real-time SLAM systems (ICRA 2020), an exoskeleton for teleoperation called DE VITO (Best Paper Award at TAROS 2019 - the UK’s largest robotics conference), a software architecture and play fetch demo for Robot DE NIRO (journal Frontiers in Robotics and AI; IROS workshop (spotlight)) and haemorrhage diagnosis with recurrent neural networks at the NeurIPS 2018 ML4H workshop (spotlight).

I studied computer science (MSc) at Imperial College London, and industrial engineering (BSc+MSc) at Karlsruhe Institute of Technology in Germany. During my degrees, I studied at and visited Tsinghua University (清华大学) in Beijing, Shanghai Jiao Tong University (上海交通大学), the University of Oxford, Singapore Management University, and Carnegie Mellon University in the US.

I enjoy playing the guitar, singing and racket sports. I speak German, English and Chinese.

Do reach out to me if my research is of interest to you - I’m always happy to chat about potential collaborations!

Contact: fabian.falck ‘at’ stats.ox.ac.uk

News

Interests
  • Probabilistic Deep Learning
  • Generative Modelling
  • Causal Inference
  • Applications in Health
Education
  • PhD in Statistical Machine Learning

    University of Oxford

  • MSc in Computer Science, 2018

    Imperial College London

  • BSc+MSc in Engineering, 2015+2016

    Karlsruhe Institute of Technology

Publications

(2023). A Unified Framework for U-Net Design and Analysis. NeurIPS 2023.

PDF Cite

(2022). A Multi-Resolution Framework for U-Nets with Applications to Hierarchical VAEs. NeurIPS 2022 (oral).

PDF Cite

(2022). Neural Score Matching for High-Dimensional Causal Inference. AISTATS 2022 (also presented at American Causal Inference Conference 2022).

PDF Cite Code

(2021). Machine Learning for Health (ML4H) 2021. ML4H 2021.

PDF Cite

(2021). Multi-Facet Clustering Variational Autoencoders. NeurIPS 2021.

PDF Cite Code

(2021). Identification of Underlying Disease Domains by Longitudinal Latent Factor Analysis for Secukinumab Treated Patients in Psoriatic Arthritis and Rheumatoid Arthritis Trials. American College of Rheumatology Convergence 2021.

PDF Cite

(2021). Ivy: Templated Deep Learning for Inter-Framework Portability. Preprint.

PDF Cite Code Project

(2020). Machine Learning for Health (ML4H) 2020: Advancing Healthcare for All. Proceedings of Machine Learning Research.

PDF Cite

(2020). Robot DE NIRO: A Human-Centered, Autonomous, Mobile Research Platform for Cognitively-Enhanced Manipulation. Journal Frontiers in Robotics and AI, also at IROS 2019 Workshop Towards Robots that Exhibit Manipulation Intelligence (spotlight & poster).

PDF Cite Code Video Video2 WorkshopPDF Documentation

(2020). Comparing View-Based and Map-Based Semantic Labelling in Real-Time SLAM. ICRA 2020 (presentation).

PDF Cite

(2019). Detecting Patterns of Physiological Response to Hemodynamic Stress via Unsupervised Deep Learning. NeurIPS 2019 ML4H Workshop.

PDF Cite

(2019). Machine Learning for Health 2019: What Makes Machine Learning in Medicine Different?. Proceedings of Machine Learning Research.

PDF Cite

(2019). Measuring Proximity between Newspapers and Political Parties – The Sentiment Political Compass. Journal Policy & Internet (presentation), also at The Internet, Policy & Politics Conference, University of Oxford.

PDF Cite

(2019). From Accuracy to Versatility: Analysing Text Classification Models regarding Transfer Learning. IEEE International Workshop on Deep and Transfer Learning.

PDF Cite

(2019). DE VITO: A Dual-arm, High Degree-of-freedom, Lightweight, Inexpensive, Passive Upper-limb Exoskeleton for Robot Teleoperation. 20th Conference Towards Autonomous Robotic Systems (Best Paper Award & presentation).

PDF Cite Code

(2018). Deep Sequence Modeling for Hemorrhage Diagnosis. NeurIPS 2018 ML4H Workshop (spotlight), also at International Symposium on Intensive Care and Emergency Medicine (abstract & poster presentation).

Cite