Topic outline

  • General

    Overview: 

    This year the course focuses on deep latent variable models for biomedical data analysis. The course starts with a background in statistical and machine learning methods for latent variable models. The main focus of the course is on methods and specific latent variable model variants for analyzing various types of biomedical data. Typical biomedical analysis tasks include e.g. exploratory and unsupervised data analysis, survival analysis, treatment effect estimation, causal analysis, time-series and longitudinal data analysis, point processes, and handling structured missingness. Lectures will cover a collection of these biomedical analysis tasks and introduce probabilistic machine learning methods for these tasks, with special focus on deep latent variable models. 

    Prerequisite knowledge: 

    Basic math courses, linear algebra, basics of probability and statistics, basics of machine learning. 

    Organisation of the course: 

    Weekly lectures in T1 on Fridays, 12:15-14:00. 

    Requirements: 

    To pass the course you need to: 

    • Attend all lectures (1 or 2 absences allowed, further absences can be compensated by completing extra reviewing duties). 
    • A project work and literature review. The literature review part involves reading recent scientific papers on a chosen topic and summarizing the articles in a review report. The project work part involves testing some of the statistical and machine learning methods on chosen datasets. The project can be done as a group of (max.) two students. 
    • Reviewing duty: review other student's literature review and project report and provide constructive feedback. 

    General information: 

    • Credits: 5 ECTS 
    • Teachers in charge: Harri Lähdesmäki 
    • Lecturers: Harri Lähdesmäki, Miika Koskinen, Mine Ögretir, Manuel Haussmann, Maksim Sinelnikov. 
    • Materials: lecture notes, parts of the book "Probabilistic Machine Learning: Advanced Topics" by Kevin Murphy (https://probml.github.io/pml-book/book2.html).

    Tentative schedule: 

    01.03.2024: Lecture 1: Background on statistical methods: divergence measures, Monte Carlo, importance sampling, multilayer perceptrons

    08.03.2024: Lecture 2: Latent variable models, variational inference, auto-encoding variational Bayes, VAEs 

    15.03.2024: Lecture 3: Generating evidence through clinical data. Electronic health records and variational autoencoders. Dr. (Tech.), Docent, Miika Koskinen, ICT Management Helsinki University Hospital

    23.03.2024: Lecture 4: Survival analysis, Mine Ögretir, Aalto University

    29.03.2024: Easter, no lecture

    05.04.2024: Lecture 5: Temporal and longitudinal data analysis 

    12.04.2024: Lecture 6: Analyzing irregularly-sampled data with structured missingness, Maksim Sinelnikov, Aalto University

    19.04.2024: No lecture (exam week)

    26.04.2024: Lecture 7: (online via zoom) Treatment effect estimation, Manuel Haussmann

    22.05: Project work and literature review report deadline 

    28.05: Review feedback deadline 

    31.05: Deadline for the final literature report