Kommentar |
This lecture will be held by Dr. Gerhardus (DLR-Institute of Data Science, Jena).
You can apply for it with the paper "Modulprüfungsanmeldung" which you can find on the faculty homepage.
According to the university's request the course will be **online during the first few weeks** (despite of the "Präsenz" in the module's title). The link to the video conference room will be communicated privately via Moodle and email after registering. In coordination with the participants we will then **switch to an in-person or hybrid format presumably in mid May**. |
Literatur |
- Peters, J., Janzing, D., and Schölkopf, B., Elements of causal inference: Foundations and
Learning Algorithms (MIT Press, Cambridge, 2017)
- Pearl, J., Glymour, M., Jewell, N. P., Causal Inference in Statistics: A Primer (Wiley, 2016)
- Pearl, J., Causality: Models, Reasoning, and Inference, 2nd edition (Cambridge University
Press, New York, 2009)
- Spirtes, P., Glymour, C., and Scheines, R., Causation, Prediction, and Search (MIT Press,
Boston, 2000)
|
Bemerkung |
Learning goals: Skills and knowledge
Conceptual understanding of the modern causal inference framework based on causal Bayesian networks and structural causal models, its enabling assumptions, typical applications, and important algorithms.
Learning goals: Abilities
Ability to frame causal questions within the causal inference framework, to select appropriate algorithms, and to interpret and communicate their results. |
Lerninhalte |
Course summary:
In this course you will learn about the modern statistical framework of "causal inference". This framework formalizes the notions of cause and effect in mathematical terms, delineates the distinction between mere statistical relationships on the one hand and causal relationships on the other hand, and provides methods that allow to learn and reason about causal relationships in a data-driven way.
Causal inference has in recent years been gaining increasing interest from the machine learning community because models informed by causal knowledge are expected to show better out-of-distribution generalization. Its methods are also applied in various other scientific fields to reason about causal relationships in cases where controlled experimentation is not feasible.
The focus of this course in on conveying the framework's conceptual basics and central concepts. It will also cover a selection of concrete algorithms and applications. A particular emphasis lies on the application of causal inference to time series data, which is ubiquitious in many applied fields.
Overview of topics:
- Distinction between statistical and causal relationships, interventions
- Causal Bayesian networks, structural causal models
- Causal effects, causal effect identification
- Causal structure learning, statistical independence tests
- Causal structure learning in time series
- Causal inference in the presence of latent variables and selection variables
- Counterfactuals
- Selected algorithms for causal effect and causal structure identification
|