Thema der Dissertation:
Biclustering Approaches Based On Correspondence Analysis For Single-Cell Transcriptomics Thema der Disputation:
Autoencoders and Variational Autoencoders for Single-Cell RNA Sequencing Data
Biclustering Approaches Based On Correspondence Analysis For Single-Cell Transcriptomics Thema der Disputation:
Autoencoders and Variational Autoencoders for Single-Cell RNA Sequencing Data
Abstract: Autoencoders (AEs) are unsupervised neural-network-based models that learn an informative latent space. Due to their non-linear nature, AEs can effectively capture complicated relationships in the data and are ideally suited for applications such as dimensionality reduction, denoising or data compression. Variational Autoencoders (VAEs) extend the autoencoder architecture to model the underlying distribution of the latent space and the data. This allows for a smooth latent space that can be sampled to generate new, unseen data and provides a flexible framework that can incorporate additional information to extend to other applications, such as batch correction. The talk will present the architecture of AEs and VAEs and how they can be trained and used for applications on single-cell RNA sequencing data.
Zeit & Ort
16.02.2026 | 14:00
Seminarraum 005
(Fachbereich Mathematik und Informatik, Takustr. 9, 14195 Berlin)