Publications
Throughout my PhD I have completed a number of projects and published these in various journals and conferences. The main focus of my research has been on the development of novel methodologies for the analysis of large scale datasets. Typically this involves the development of new algorithms and software, which often give state of the art performance on a range of benchmark datasets.
The following is a list of my publications to date, also available on my Google Scholar profile.
2024
-
Logistic Variational Bayes Revisited
Michael Komodromos, Marina Evangelou, and Sarah Filippi
In Forty-first International Conference on Machine Learning (ICML) , 2024
A novel variational Bayes algorithm for binary classification models is developed. The algorithm improves upon existing methods by providing a more accurate approximation to the posterior distribution. In real terms, this leads to improved classifcation accuracy and improved uncertainty quantification.
2023
-
Group spike-and-slab variational Bayes
Michael Komodromos, Kolyan Ray, Marina Evangelou, and Sarah Filippi
2023
Under review
We introduce a novel variational Bayes algorithm for high-dimensional GLMs where a group strcutre is present. Our algorithm is orders of magnidute faster than existing Bayesian methods (typically 200 - 300 times faster), and provides an accurate posterior approximation. Usecases include gene expression data and other high-dimensional non-parametric models.
2022
-
