Besides research and communication, I am an enthusiastic advisor and educator (primarily for machine learning, but I’m occasionally trying to expand into other areas of computer science and mathematics).
The best way to understand something is to explain it to somebody else.
Guided by this quote, I have readily taken up teaching, lecturing and advising throughout my PhD, and am still keen to advise interesting projects whenever they arise.
Alongside Pietro Liò, I am co-teaching the Representation Learning on Graphs and Networks (L45) Master’s course at Cambridge. This is an extended version of the Graph Neural Networks module for the Advanced topics in machine learning or natural language processing (R250) course, which we have delivered in the past.
Alongside Michael Bronstein, Joan Bruna and Taco Cohen, I have delivered the Geometric Deep Learning (GDL100) Master’s course for the African Master’s in Machine Intelligence. This course follows our recently released proto-book and we have made all artifacts and resources publicly available! Since then, I have given an adapted version of this course for the 47th Heidelberg Physics Graduate Days.
I have, on several occasions, also held formal (guest) lectures to undergraduate and postgraduate students:
Geometric Deep Learning: GNNs Beyond Permutation Equivariance. Machine Learning on Graphs (CS224W), Stanford University, 30 November 2021.
Everything is Connected: Graph Neural Networks from the Ground Up. Deep Learning Course (UVADLC), Universiteit van Amsterdam. 26 November 2021.
Neural Algorithmic Reasoning. Network Data Science 2 (PHYS 7332), Northeastern University, 2 April 2021.
Be nice to your neurons: Initialisation, Normalisation, Optimisation and Regularisation. Introduction to Deep Learning (COMPGI23), University College London, 14 November 2017.
Unsupervised methods: Diving deep into autoencoders. Introduction to Deep Learning (COMPGI23), University College London, 10 October 2017.
A trip down long short-term memory lane. Research Students Lecture Series, University of Cambridge, 21 February 2017.
Furthermore, I have been in charge of designing the practical session on neural networks for the Cambridge MPhil course on Machine Learning and Algorithms for Data Mining (L42).
While at Cambridge, I have been extremely fortunate to (co-)supervise the project work of several students, whose dissertation titles are outlined below. Most of these disseratations have been turned into research publications, and many of the students are now working in prominent AI labs (either as doctoral students or research engineers).
- Ioana Bica (2017) Comparative Analysis of Neural Network Architectures for Epigenetics Inference.
- Momchil Peychev (2017) Experimental Study on the Properties of Disentangled Autoencoders.
- Tudor Petru Țiplea (2017) An end-to-end control policy system for the GameBoy platform. (Highly commended dissertation)
- Andreea-Ioana Deac (2018) Neural architectures for end-to-end paratope prediction.
- Aliyah Bond (2019) Intelligent navigation of text adventure games.
- Carlos Purves (2019) The PlayStation reinforcement learning environment.
- Benjamin Day (2016) Combining Machine Learning & Multilayer Network Theory.
- Edgaras Liberis (2017) Antibody Binding Residue Prediction.
- Laurynas Karazija (2017) Automatic inference of cross-modal connection topologies for X-CNNs. (Dr John Maheswaran Prize for a Highly Commended Project Report)
- Cătălina Cangea (2017) 1D-2D Cross-modality for deep audiovisual classification.
- Miloš Stanojević (2018) Neural network hyperparameter optimisation on small validation sets.
- Nathaniel McAleese-Park (2018) Cross-modal learning for the joint understanding of 2D and 3D data.
- Aaron Solomon (2019) Dynamic temporal analysis for graph structured data.
- Felix Opolka (2019) Representation learning for spatio-temporal graphs.
I serve as reviewer for several major machine learning conferences, journals and workshops, always striving to provide the most actionable feedback possible. My efforts have been recognised with best reviewer awards at all of the “big three” machine learning conferences: NeurIPS 2019, ICML 2020 and ICLR 2021.
- Neural Information Processing Systems (NeurIPS) 2019, 2020, 2021
- International Conference on Machine Learning (ICML) 2020
- International Conference on Learning Representations (ICLR) 2020, 2021, 2022
- The AAAI Conference on Artificial Intelligence (AAAI) 2022
- International Joint Conference on Artificial Intelligence (IJCAI) 2021
- International Joint Conference on Neural Networks (IJCNN) 2019
- European Symposium on Artificial Neural Networks (ESANN) 2018
- Nature Communications
- Nature Machine Intelligence
- Communications Medicine
- Journal of Machine Learning Research (JMLR)
- IEEE Computer
- Wordplay: When Language Meets Games Workshop at NeurIPS 2020
- Deep Learning on Graphs: Methods and Applications (DLG) Workshop at KDD 2020
- Bridge Between Perception and Reasoning: Graph Neural Networks & Beyond Workshop at ICML 2020
- Learning and Reasoning with Graph-Structured Representations Workshop at ICML 2019
- Representation Learning on Graphs and Manifolds (RLGM) Workshop at ICLR 2019
Public policy 🇷🇸
I serve on the Advisory Board of the Serbian AI Society, an organisation aiming to develop and maintain a plan for the improvement of the Serbian AI ecosystem in the coming years.
I serve on the Scientific Advisory Board of the Institute for AI Research and Development of Serbia.
I’ve participated in developing Serbia’s AI Strategy for 2020-2025, as one of the representatives of Serbian machine learning researchers working abroad. The Government of Serbia had officially adopted it on 26 December 2019.
The Strategy defines objectives and measures for the development of AI, whose implementation should result in economic growth, improvement of public services, advancement of scientific staff, and development of skills for the jobs of the future. Also, the implementation of the Strategy measures should ensure that AI in the Republic of Serbia is developed and implemented in a safe manner and in accordance with internationally recognised ethical principles, in order to exploit the potential of this technology to improve the quality of life of each individual and the society as a whole, as well as to achieve Sustainable Development Goals.