Extreme analyses of graphs and text
Prof. Dr. Ansgar Scherp, Depart. of Data Base and Information Systems, University of Ulm
| Kurzübersicht | |
|---|---|
| Art des Termins |
|
| Wann |
05.07.2022 von 11:00 bis 12:00 |
| Wo | Room no. 0.016 |
| Termin übernehmen |
 vCal iCal
|
Scientific talk: Extreme analyses of graphs and text
This talk presents my research on extreme analyses of graphs and text. For example, in text analysis we consider the task of extreme multi-label classification and show that our simple WideMLP outperforms state-of-the-art graph-based models like TextGCN. In graph analysis, we are analysing graphs with billions of edges. Our CIKM 2020 paper shows that graph summaries can be efficiently computed in a parallel and incremental algorithm. This is important for reflecting temporal changes in web graphs. Finally, we are considering my most recent research on stratified k-bisimulation on very large graphs with up to almost two billion edges.
Teaching talk: Logistic Regression: Classification
Linear and non-linear regression are well-known approaches for modeling data distributions. In the MSc module "Data Mining and Machine Learning", we are considering logistic regression to effectively solve binary classification tasks. We show how the maximum likelihood is computed for logistic regression and why it is a strong non-neural model for tasks such as text classification. As subsequent method, we would see multi-layer perpectron models (see WideMLP in the research talk!), which forms the foundation of modern deep learning.
Location (for both talks): Institute of Computer Science, Friedrich-Hirzebruch-Allee 8, 53115 Bonn, Room no. 0.016
