2 PhD Students at Centrum Wiskunde & Informatica

The Dutch national research institute for mathematics and computer science (CWI, Centrum Wiskunde & Informatica) located in Amsterdam, and the Leiden University Medical Center (LUMC) together have 2 vacancies for fully funded

PhD students,

on a joint project on the development of evolutionary machine learning techniques for explainable AI and their application to medical data.

Job description
CWI closely collaborates with LUMC to work on innovations in the medical domain along the entire spectrum from algorithmic foundations to clinical integration. Currently, CWI (the Life Sciences and Health research group) and LUMC (the department of radiation oncology) have a joint project for which we seek 2 talented PhD students (one to be hired at CWI and one to be hired at LUMC) to work together on new techniques for and novel combinations of evolutionary algorithms and machine learning, in a project funded by the Dutch Research Council (NWO, Nederlandse Organisatie voor Wetenschappelijk Onderzoek) entitled "EXAMINE - Evolutionary eXplainable Artificial Medical INtelligence Engine".

Machine learning (ML), especially deep learning, has shown to be capable of impressive results in many fields of science and society. Nevertheless, in most cases, learned models are still incomprehensible black boxes for humans. Especially for disciplines such as medicine, this is a bottleneck for widespread adoption, because for clinical use, medical experts want to understand what has been learned and why.

We will therefore focus on types of ML that provide models that can be insightfully inspected and understood. Specifically, these models are made up of (combinations of) symbolic expressions, deep neural networks for medical images and radiation dose distributions (used for treating cancer), and Bayesian networks that can be used to create explainable prediction models (e.g., of treatment outcomes).

To learn such explainable models, powerful and versatile optimization algorithms are required. For this, we will innovate the recently introduced state-of-the-art research line on model-based evolutionary algorithms known as the Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA) family and exploit recently published new insights into its (GPU) parallelization capabilities.

The intention is that both PhD students will work on novel combinations of evolutionary algorithms and machine learning. However, one PhD student is intended to focus more on symbolic expression evolution and the other on deep neural network evolution. The students are furthermore expected to work together on integrating their results with Bayesian network evolution and to perform clinical validation studies using data of gynaecological cancer patients. Candidates are therefore expected to spend at least one day per week at the other institute.

Further details:
2 PhD Students at Centrum Wiskunde & Informatica