About Me

I am a machine learning engineer and researcher. Currently, I am ending my PhD at ETH Zürich. My main research interest is deep generative modeling – see our new SDN-VAE for image modeling. I am also enthusiastic about applying machine learning to biology – see our web platform for automatic analysis of sleep patterns with over 10.000 submissions worldwide.

At ETH, I work with Prof. Joachim. M. Buhmann. I collaborated with Swiss AI lab IDSIA led by Prof Jürgen Schmidhuber and with the Max Planck Institute for Intelligent Systems led by Prof. Bernhard Schölkopf, to which I am affiliated. I also worked with the group of Prof. Steven Brown in interdisciplinary collaboration. Previously, I worked in Logitech in Lausanne and Disney Research in Zürich with Dr. Paul Beardsley and Prof. Otmar Hilliges. I obtained my MS from ETH Zürich and my BS from the University of Belgrade, both in computer science.

I was born in Belgrade, Serbia in 1990. The English spelling of my name is Djordje Miladinovic. In Serbian, it would be Đorđe Miladinović in Latin and Ђорђе Миладиновић in Cyrillic alphabet. If you can pronounce Django or Djokovic, then you can also pronounce Djordje.

Research Highlights

The first line of my research is deep generative modeling with a particular focus on variational autoencoders and deep autoregressive models. I explore techniques for learning probabilistic models (density estimation) and discovering 'useful' and structured representations in an unsupervised fashion.

The second line of my research is the application of deep learning to biological and medical data. I am particularly intrigued by sleep -- humans spend roughly one-third of their lives sleeping and yet so little is known about this mysterious phenomenon.

Teaching & Mentoring

During my time at ETH Zurich, I was involved in the organization and teaching of the following courses:

I also supervised a number of master students on various topics:


Spatially Dependent U-Nets: Highly Accurate Architectures for Medical Image Segmentation
João B. S. Carvalho, João A. Santinha, Ðorđe Miladinović, Joachim M. Buhmann
arXiv preprint arXiv:2103.11713
[Link to paper] [bibtex]

Dynamic Dropout: Regulating Teacher Forcing in Autoregressive Models
Ðorđe Miladinović and Joachim M. Buhmann
Under review.

Spatial Dependency Networks: Neural Layers for Improved Generative Image Modeling
Ðorđe Miladinović, Aleksandar Stanić, Stefan Bauer, Jürgen Schmidhuber, and Joachim M. Buhmann
To appear in the 9th International Conference on Learning Representations, ICLR 2021.
[Link to paper] [bibtex]

Instantaneous Metabolic Changes with Sleep Stage Transitions Observed in Exhaled Breath
Nora Nowak, Thomas Gaisl, Ðorđe Miladinović, Ričards Marcinkevičs, Martin Osswald, Stefan Bauer, Joachim M. Buhmann, Renato Zenobi, Pablo Sinues, Steven Brown, Malcolm Kohler.
Under review.

On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset
Muhammad Waleed Gondal, Manuel Wuthrich, Ðorđe Miladinović, Francesco Locatello, Martin Breidt, Valentin Volchkov, Joel Akpo, Olivier Bachem, Bernhard Schölkopf, and Stefan Bauer.
In Advances in Neural Information Processing Systems, pages 15740–15751, NeurIPS 2019.
[Link to paper] [bibtex]

Disentangled State Space Representations
Ðorđe Miladinović, Muhammad Waleed Gondal, Bernhard Schölkopf, Joachim M Buhmann, and Stefan Bauer
arXiv preprint, arXiv:1906.03255, 2019 & DeepGenStruct Workshop at ICLR 2019.
[Link to paper] [bibtex]

Robustly Disentangled Causal Mechanisms: Validating Deep Representations for Interventional Robustness
Raphael Suter, Ðorđe Miladinović, Bernhard Schölkopf, and Stefan Bauer.
In International Conference on Machine Learning, pages 6056–6065. PMLR, ICML 2019.
[Link to paper] [bibtex]

SPINDLE: End-to-end learning from EEG/EMG to extrapolate animal sleep scoring across experimental settings, labs and species
Ðorđe Miladinović, Christine Muheim, Stefan Bauer, Andrea Spinnler, Daniela Noain, Mojtaba Bandarabadi, Benjamin Gallusser, Gabriel Krummenacher, Christian Baumann, Antoine Adamantidis, Steven A. Brown , Joachim M. Buhmann.
PLoS computational biology 2019.
[Link to paper] [bibtex]

Granger-Causal Attentive Mixtures of Experts: Learning Important Features with Neural Networks
Patrick Schwab, Ðorđe Miladinović, and Walter Karlen.
In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, pages 4846–4853, AAAI 2019.
[Link to paper] [bibtex]

Efficient and Flexible Inference for Stochastic Systems
Stefan Bauer, Nico S Gorbach, Ðorđe Miladinović, and Joachim M Buhmann.
In Advances in Neural Information Processing Systems, pages 6988–6998, NeurIPS 2018.
[Link to paper] [bibtex]