Deep learning in biomedicine

The goal of this course is to cover recent deep learning methods and demonstrate how they can be applied to biomedical data. The course will cover ongoing advances in deep learning research for different input data types (e.g., convolutional neural networks for images, graph convolutional neural networks for graph structured data, transformers for sequence data). We will start with a standard supervised learning setting and then cover the ongoing developments in methodologies that allow us to learn using scarcely labeled datasets by transferring knowledge across tasks (e.g., transfer learning, meta-learning). These settings have particular importance in the biomedical domain in which it is often very difficult to obtain labeled datasets. Recent papers from the literature that apply these methods to problems in biomedicine will be presented and discussed.
In assignments, students will work with popular deep learning software frameworks. They will be evaluated on their ability to understand and implement the methods learned in a class. In the project, students will choose a real-world problem in the biomedical domain and develop a solution for the problem of their choice. They will be evaluated on the ability to propose and develop a suitable model to solve the task, propose suitable evaluation, provide analysis and extract insights from the developed models, write a project report and present project results.
This course is of interest to MS/PhD students interested in recent deep learning methods and their applications to real-world problems in the biomedical domain.