Solving various complex challenges (i.e., reducing methane emissions or reducing the infection rate during a pandemic) requires understanding how changes to a complex system influence outcomes. This knowledge helps identify interventions that can be implemented to achieve specific goals.

Automating the process of answering interventional questions (e.g., which action to take and to what extent) will help decision-makers solve such challenges. The idea of this project lies at the core of the vision of automated intervention development based on spatio-temporal datasets generated by modern sensing technologies. Most machine learning algorithms designed for spatio-temporal do not allow assessing the impact of interventions without capturing causal links.

At the same time, configuring causal machine learning algorithms is highly complex, being an unsupervised learning task. We aim to address the core novel objective of scalably assessing the impact of possible interventions through counterfactual prediction based on spatio-temporal data.

• algorithms
• complex systems

Share on Facebook Share by Bluesky Share on LinkedIn Share by WhatsApp Share by Mastodon