BLOGS
Advancing AI Frontiers: The ELLIS Amsterdam MSc Honours Programme (2/2)
The ELLIS unit in Amsterdam is dedicated to developing and retaining a new wave of AI researchers through its partnerships, which are crucial for achieving scientific distinction and pioneering research in modern AI across Europe. To achieve this, the ELLIS unit created a MSc Honours programme in 2020.
The goal of the programme is to enable excellent MSc students to have their AI thesis projects supervised by ELLIS researchers in different countries. The programme therefore does not only ensure diverse and constructive feedback from researchers of different institutions, but also gives the students the opportunity to expand their network within their area of interest and experience different research environments. Finally, working with high-level researchers enables the students to research state-of-the-art AI topics and be on the front of pushing the boundaries of AI-research in Europe.
We interviewed 13 MSc Honours students and highlighted how the programme helped them reach their academic goals, their key takeaways from the programme, and what is next for them. This post will cover the second half of the students.
We would like to congratulate all of the ELLIS unit Amsterdam’s MSc Honours students for their exceptional achievements!
Übersetzt mit DeepL.com (kostenlose Version)
David Frühbuss
Thesis project title: Cancer Immunotherapy Design with Geometric Deep Learning
Supervisors:
– Dr. Erik Bekkers (ELLIS Member, University of Amsterdam)
– Prof. Dr. Li Xue (Radboud University)
– Prof. Dr Stefanie Jegelka (ELLIS Member, TUM München)
Research Summary:
The main goal of this thesis is to build models that can predict whether a molecule is likely to bind with a certain protein based on their joint predicted 3D structure. David and his supervisors focus on peptides and a specific protein called Major histocompatibility complex (MHC). Next to T-Cells, this protein is the most important component in the body’s immune response to cancer cells. Being able to reliably predict which peptides bind well to this protein would enable new and improve existing cancer therapies.
MSc Honours Opportunity:
The ELLIS Honours programme gives me the unique opportunity to work with two of the best researchers in Geometric Deep Learning. Together with Li Xue, who is an expert in applying AI to cancer research, this combination of expertise allows us to push what is possible in designing cures for various forms of cancer. In my upcoming research visit to the Jegelka Lab and through the ELLIS network in general I hope to foster many research connections across Europe, which will hopefully allow me to further pursue research like this in the future.
Veljko Kovac
Thesis project title: Higher Order Graph Neural Networks
Supervisors:
– Dr. Erik Bekkers (ELLIS Member, University of Amsterdam)
– Prof. Dr. Pietro Liò (Ellis Member, University of Cambridge)
– Dr. Petar Veličković (ELLIS Scholar, University of Cambridge, Google DeepMind)
– MSc Floor Eijkelboom (University of Amsterdam)
Research Summary:
Incorporating higher order information is crucial for enhancing our comprehension of complex systems across various fields. For example, in protein-protein interaction networks, cliques indicate protein complexes, highlighting the importance of moving beyond pairwise interactions. However, the shift from analyzing basic pairwise interactions to harnessing the full potential of higher order information introduces substantial challenges, especially in identifying and leveraging useful higher order structures; a process often guided by domain-specific knowledge. The complexity of choosing the right substructures due to combinatorial issues is a significant obstacle. This project focuses on developing methods to efficiently identify and leverage these higher order structures within graphs, aiming to improve the understanding of complex systems in various domains.
MSc Honours Opportunity:
As part of the ELLIS MSc Honours program, I am given the opportunity to conduct a research visit at my co-supervisor’s institution for a month. Stationed at the University of Cambridge and collaborating with highly skilled researchers in my field, I am confident that this experience will not only advance my project significantly but also expose me to various other intriguing projects currently in progress. Additionally, this visit presents a fantastic chance to meet people and expand my professional network, which is an aspect I am particularly excited about. I feel incredibly fortunate to have this opportunity and am fully committed to making the most out of it.
Diego Garcia Cerdas
Thesis project title: Data-driven exploration of selectivity in the human visual cortex using diffusion models
Supervisors:
– Dr. Iris Groen (ELLIS Member, University of Amsterdam)
– Prof. Dr. Gemma Roig (ELLIS Member, Goethe University Frankfurt)
– Dr. Pascal Mettes (ELLIS Member, University of Amsterdam)
– MSc Christina Sartzetaki (University of Amsterdam)
Research Summary:
The availability of large-scale neuroimaging datasets has opened new possibilities for studying the brain in a data-driven manner. By analyzing brain responses to thousands of naturalistic image and video stimuli, it is possible to discover cortical organization properties possibly overlooked by previous hypothesis-driven research. These properties comprise, for example, regions in the higher visual cortex that are selectively engaged in perceiving certain objects like faces, bodies, places, words, or food.
This thesis project explores how one can visually understand the properties suggested by these large datasets using deep generative models. Particularly, diffusion models can be used to generate stimuli that are predicted to maximally activate a targeted brain region. However, the fidelity of these results will depend on the type of representation we extract from the stimuli to predict brain activations. Therefore, a systematic study of different design choices is needed before one can extract relevant and truthful conclusions from the data.
MSc Honours Opportunity:
Pursuing the MSc Honours programme has been a great justification for challenging myself and exploring the implications of AI in broader research areas. Most importantly, I have been able to do so with the close guidance of excellent professors and PhD students at the UvA. I am excited to visit Goethe University Frankfurt in the coming months, and I expect it to be an enriching experience both personally and professionally. I am confident that the results of this Honours programme will open many possibilities and prepare me for pursuing a PhD after my Masters.
Luca Petru Pantea
Thesis project title: Finetuned Spatiotemporal Graph Neural Networks for Contact Tracing
Supervisors:
– MSc Rob Romijnders (ELLIS Member, University of Amsterdam)
– Dr. Yuki Asano (ELLIS Member, University of Amsterdam)
– Prof. Pascal Frossard (ELLIS Fellow, EPFL Lausanne)
Research Summary & MSc Honours Opportunity:
The research enhances statistical contact tracing by integrating spatiotemporal data and applying Reinforcement Learning for finetuning, aiming to develop more effective COVID-19 testing policies. It leverages decentralized Graph Neural Networks to improve infection risk estimation, seeking to reduce peak infection rates and hospitalizations.
MSc Honours Opportunity:
Through the ELLIS MSc Honours Programme, I’m eager to dive into research that has the potential for real-world impact, while focusing on a topic I’m deeply passionate about: Graphs. The programme offers a unique opportunity to work in various research environments at UvA and EPFL, fostering a proactive atmosphere that nurtures the exchange and development of ideas. This allows me to interact with and learn from leading academics, while additionally providing me with essential resources and guidance, enabling me to make a meaningful contribution to the field.
Oline Ranum
Thesis project title: Geometric Sign Language Processing
Supervisors:
– Dr. Erik Bekkers (ELLIS Member, University of Amsterdam)
– Prof. Dr. Coloma Ballester (ELLIS Member, University of Barcelona)
– Prof. Dr. Floris Roelofsen (University of Amsterdam)
Research Summary:
Warm AI is the pursuit of technologies that assist people living under pressing social conditions. A prime example is sign language processing (SLP), which automates interpretations between d/Deaf and hearing communities. In this thesis project, Oline and her supervisors contribute to improving sign language recognition (SLR) by conditioning neural networks on the linguistic structures of signs. This includes embedding sign geometry as inductive biases via GNNs and making use of spatial symmetries to encourage view-invariant SLR. To achieve these goals Oline has developed a novel pipeline for data production via motion capture and semi-automatic annotation methods of phonological features. Her method is called geometric sign language processing, as it builds on an underlying principle of informing representations with geometric structures that naturally emerge from the visual modality of signing.
MSc Honours Opportunity:
As an MSc Honours Student, I have the opportunity to visit Barcelona, a vibrant city for sign language recognition and computer vision research. The Intelligent Multimodal Vision Analysis group of Dr. Ballester can offer insights complementary to the resources available at the UvA, which can be greatly beneficial towards advancing geometric SLP. I hope to use my visit to connect with other researchers in my field, introduce and discuss geometric SLP and learn more about alternative SLR approaches under development in Barcelona!
Marten Turk
Thesis project title: Using graph neural networks to predict people’s infectiousness under conditions of low adoption on the contact tracing apps
Supervisors:
– Prof. Dr. Max Welling (ELLIS Fellow, University of Amsterdam)
– Prof. Dr. Ralf Herbrich (ELLIS Member, University of Potsdam)
– MSc Rob Romijnders (ELLIS PhD, University of Amsterdam)
Research Summary:
The global impact of COVID-19, causing trillions of dollars in economic losses and around 7 million deaths, pushed governments to implement strategies to mitigate the disease. One such strategy was the use of contact tracing, where phones can exchange messages to notify users of potential exposure to the virus. Soon after the pandemic began, researchers started to develop AI methods to effectively mitigate the pandemic. These methods outperformed the simpler ones used before, however most of them assumed full adoption rates of the contact tracing apps. This is where the research by Marten and his supervisors stands out, as their goal is to develop an applicable method using graph neural networks to predict people’s infectiousness under conditions where the adoption rate of the apps is below 60%. They have chosen this rate of adoption as this is the point where baseline methods started to struggle.
MSc Honours Opportunity:
Being a part of the ELLIS MSc Honours program is a great opportunity to foster collaborations between different institutions and meet talented researchers. I plan to make the most of it by going to the Hasso Plattner Institute for the whole month of April as a visiting researcher to network, collaborate and learn with the researchers in the Artificial Intelligence and Sustainability research group. I am excited for the collaboration and hope to make the most of it as I progress in my work.
See more information regarding the MSc Honours programme.