Adjoint scientifique HES (postdoc)|
Amina Mollaysa joined the DMML team as a research and teaching assistant in March 2015. After finishing her bachelor on Theoretical Math, she is leading her carrier where the theory and application meet. She has studied MathMods - Mathematical Modelling in Engineering: Theory, Numerics, Application Master Program at the University of Nice Sophia Antipolice, University of Hamburg and the University of L'Aquila. She has worked on the free-flight routing problem for her master thesis within the E-Motion project in Hamburg. She has defended her Ph.D. thesis on 2021 February in the Department of Computer Science at the University of Geneva under the joint supervision of Profs. Alexandros Kalousis and Stephane Marchand-Maillet. After defence, she has continued working in the DMML ream as a postdoctoral researcher.
The main focus of her research are representation learning, generative models for discrete data, and optimization algorithms. Specifically, she has worked on how to incorporate meta-features (feature side-information) into learning to improve generalization performance, and how to enable the conditional generation and style transfer over discrete structured data.
Currently, she is particularly interested in generative models over discrete structured data, such as molecules, trees, or graphs. These are typically hard to optimize, especially for the conditional generation. Methods developed for natural images often fail dues to the discrete nature of the space. She is interested in developing efficient algorithms that can deal with the discrete nature of the data and be able to do conditional generation and style transfer over discrete structures.
Proceedings of the 44th Conference on Neural Information Processing Systems, NeurIPS , 2020.
Proceedings of The Eleventh Asian Conference on Machine Learning, 2019.
Proceedings of the 34th International Conference on Machine Learning, ICML 2017, Sydney, NSW, Australia, 6-11 August 2017, pp. 2508–2517, 2017.