B1738
Title: Algebraic and combinatorial methods for causal model representation and selection
Authors: Liam Solus - KTH Royal Institute of Technology (Sweden) [presenting]
Abstract: When a statistical model is defined by a family of polynomial constraints, such as a graphical model or a more general conditional independence model, tools from algebraic geometry and combinatorics can allow us to prove theorems relevant in model representation and selection. The resulting insights from these theorems can also direct us towards methods for model selection, such as algorithms for data-driven learning of causal models under assumptions weaker than the classic conditions, such as faithfulness. We will present examples of such algebraic techniques and discuss their consequences in regard to the problem of causal model selection.
