Hasso-Plattner-Institut
HPI Digital Health Cluster
 

Applied Machine Learning for Digital Health

General Information

  • You are viewing an archived version of the course, please find a newer version here
  • Teaching staff: Florian Borchert, Aadil Rasheed, Dr. Mozhgan Bayat, Dr. Matthieu-P. Schapranow
  • Location: Campus III, G1-E.15/16 (The seminar is planned in presence respecting current hygenic measures to fight the COVID-19 pandemic, potentially selected online slots)
  • 4 Semesterwochenstunden (SWS) 6 ECTS (graded)
  • Limit: Max. number of participants defined by the number of provided topics.
  • Dates & times: Tuesdays and Thursdays 1.30pm (s.t.)
  • Kickoff courses: Tue Apr. 26, 2022 and Thu Apr 28, 2022 at 1.30pm (s.t.)
  • After the kickoff event (first course) you have to send us your preferred seminar topics (due date will be mentioned in the slides). Afterwards, you will be assigned to one of your preferred topics, which needs to be confirmed through official course enrollment by you.

News

  • Topics and selection procedure will be presented during the kickoff event.

Topics

  • A: A Generative Model for Resolving Coordination Ellipses in German Medical Text
  • B: Weak Supervision for Biomedical Information Extraction with skweak
  • C: Debugging NLP Models for Information Extraction from Medical Case Reports
  • D: Analysis of Censored Data Sets
  • E: Patient Survival Analysis
  • F: Feature Engineering and Hyperparameter Optimization for Clinical Prediction Models
  • G: Data Imputation Methods for Clinical Prediction Models
  • H: Identification of Patient Cohorts using Population Parameters 
  • I: Identification of Patient Cohorts using Clinical and Lab Values

Scope of the seminar

The purpose of this seminar is to help you to broaden your expertise in Machine Learning (ML) and Artificial Intelligence (AI) and apply it to selected real-world use cases.

Therefore, we will introduce selected ML/AI technologies and tools to you, which are relevant for your chosen seminar projects. You will acquire hands-on experience with these tools and apply them to real-world scenarios on realistic data sets. Please bear in mind: to allow you access to real-world data, some of the data sets might require you to either sign-up on a webpage, agree to follow data handling steps, sign a data use or confidentially agreement, or similar aspects. We will equip you with the required ML/AI expertise and provide you access to materials for your chosen projects. We expect you to deep dive in the required ML/AI technology, to do research on related work in the specific field, to design and apply your own ML/AI approach, and to evaluate your approach and compare it to results from related work. As a result, you will broaden your ML/AI skills on a real-world digital health use case, apply selected ML/AI methods, and evaluate and interprete your obtained results.

You will select your project preference from a list of seminar topics presented in the kick-off event. We will coach you throughout the whole semester with regards to the chosen project, i.e. you will have regular appointments with your tutor. Furthermore, we will provide guidance for improving your research and presentation skills throughout the seminar. Therefore, you will share your results in an intermediate and a final presentation with all participants. The presentation will help you to communicate your approach and intermediate results as well as to receive individual feedback on the approach and progress. Ultimately, you will document your findings in a scientific report at the end of the seminar.

Grading

The final grading will be determined by the following individual parts, each of them must be passed individually: 

  • Seminar results, i.e. intermediate + final presentation conducted during the seminar slots + research prototype  (40%) 
  • Research article submitted by the end of the seminar (40%)
  • Individual commitment (20%)