PhD position in evolutionary search for adaptable and interpretable machine Full-time NEW

Large, deep neural networks have demonstrated significant success in effectively distilling information from massive datasets. However, their effectiveness is limited by several drawbacks, including being data- and resource-intensive, often being non-interpretable, and being highly brittle towards shifts in data distribution and changes in the problem domain. This fully-funded three-year project, with the possible further extension to four years, will focus on leveraging the next generation of evolutionary algorithms to evolve efficient, robust, and interpretable predictors and data processing algorithms. A crucial aspect will be maintaining a pool of diverse, complementary models that can be adapted to new problems through continuous learning. We will develop this methodology to address several applications within the life sciences, with a special interest in computer vision applications, with a particular emphasis on microbial resistance. This project will be done at the KERMIT research unit at Ghent University.

Tagged as: Computer Science