TU Delft Bioengineering Institute Best MSc Graduate 2021 Akash Singh

Akash Singh won the 2nd place as the TU Delft Bioengineering Institute Best MSc Graduate 2021 for his thesis “Unsupervised Manifold Alignment with TopoGAN“.

In his thesis, Akash developed TopoGAN, a deep learning method to solve the challenging task of integrating single cell datasets with no matching samples (i.e. cells) or features. Akash’s thesis proposes multiple innovative ideas to address this challenge. First, Akash showed that Topological Autoencoders can capture the heterogeneity of single cell data better than current approaches such as (variational autoencoders, tSNE and UMAP). This on its own is a significant contribution to the field. Second, Akash proposed an approach to tackle the instability of GAN methods in the task of manifold alignment, which can be generally applied in other fields of machine learning. Third, in evaluating the performance of his method, Akash showed that current strategies have severe shortcomings and should as such be revised to faithfully reflect the performance of different methods.

Read more here.