|
Research Experience
Constraint-based structure learning based on the conditions of faithfulness
I had the honor to conduct a research project under the supervision of Dr. Kayvan Sadeghi from University College London and Dr. Javad Ebrahimi from Sharif University of Technology since July 2021 until January 2023. This was my first reseach experience and we were focused on the theoretical aspects of developing a new constraint-based structure learning algorithm based on the conditions of faithfulness. In this study, we introduced a new algorithm
that tries to form a causal graph from an input probability distribution. This algorithm is also to create a graph within some general
classes like CMGs and AnGs, but most of the results in this study
were specialized for the case of having the input distribution faithful
to a DAG. We attempted to show that the algorithm’s output is Markov
equivalent to the true causal graph that is faithful to the underlying
distribution, and the results were published as my bachelor's thesis report.
Open problems in causal structure learning: A case study of COVID-19 in the UK
In this project, I was mainly focused on the applied aspect of causal inference. I mostly used Bnlearn and Bayesys packages to implement different score-based and constraint-based algorithms (such as HC, PC-Stable, TABU, MMHC, SaiyanH, etc.) on our COVID-19 dataset and analyze the results to help us unveil the underlying causal structure. Before this step, I helped the team to discretize some variables using K-means clustering and come up with discretized versions of our dataset as well, in order to be able to use more algorithms that were only applicable to discrete data. The results were published as a paper with the same title in the journal of Expert Systems with Application (more info available in my CV)
|