In the wake of the revolution brought by Deep Learning, we believe neural networks can be leveraged as a tool in the service of dimensionality reduction (DR) for understanding large datasets with many dimensions (measurements). In this work, we present techniques for DR based on neural networks which improve over existing techniques on criteria such as scalability, dealing with unseen data, cluster separation, and ease of use, to name a few. We also present a quantitative evaluation of popular techniques, and propose novel applications that highlight the importance of DR techniques as tools for high-dimensional data analysis.
|Qualification||Doctor of Philosophy|
|Place of Publication||[Groningen]|
|Publication status||Published - 2021|