Die Dauer dieses Termins beträgt derzeit
95 Minuten. Bitte ggf. einen weiteren Raum reservieren und den Termin auf zwei Räume aufteilen. Dazu unter
Termine eine zusätzliche Terminseite anlegen und die Vorträge neu zuweisen.
Termin in Kalender importieren: iCal (Download)
Vorträge
| Vortragende(r)
|
Tobias Hombücher
|
| Titel
|
Generalized Monte Carlo Dependency Estimation and Anytime Supervised Filter Feature Selection
|
| Vortragstyp
|
Masterarbeit
|
| Betreuer(in)
|
Edouard Fouché
|
| Vortragssprache
|
|
| Vortragsmodus
|
online
|
| Kurzfassung
|
Dependency estimation is an important problem in statistics and is applied frequently in data science. As modern datasets can be very large, dependency estimators should be efficient and leverage as much information from data as possible. Traditional bivariate and multivariate dependency estimators are only capable to estimate dependency between two or n one-dimensional datasets, respectively. In this thesis, we are interested in how to develop estimators that can estimate the dependency between n multidimensional datasets, which we call "generalized dependency estimators".
We extend the recently introduced methodology of Monte Carlo Dependency Estimation (MCDE), an effective and efficient traditional multivariate dependency estimator. We introduce Generalized Monte Carlo Dependency Estimation (gMCDE) and focus in particular on the highly relevant subproblem of generalized dependency estimation, known as canonical dependency estimation, which aims to estimate the dependency between two multidimensional datasets. We demonstrate the practical relevance of Canonical Monte Carlo Dependency Estimation (cMCDE) by applying it to feature selection, introducing two methodologies for anytime supervised filter feature selection, Canonical Monte Carlo Feature Selection (cMCFS) and Canonical Multi Armed Bandit Feature Selection (cMABFS). cMCFS directly applies the methodology of cMCDE to feature selection, while cMABFS treats the feature selection problem as a multi armed bandit problem, which utilizes cMCDE to determine relevant features.
|
| Vortragende(r)
|
Jonas Zoll
|
| Titel
|
Injection Molding Simulation based on Graph Neural Networks (GNNs)
|
| Vortragstyp
|
Bachelorarbeit
|
| Betreuer(in)
|
Daniel Ebi
|
| Vortragssprache
|
|
| Vortragsmodus
|
in Präsenz
|
| Kurzfassung
|
Numerical filling simulations are an important tool for the development of injection molding parts. Existing simulations rely on numerical solvers based on the finite element method. These solvers are reliable and precise, but very computationally expensive even on simple part geometries.
In this thesis, we aim to develop a faster injection molding simulation based on Graph Neural Networks (GNNs) as a surrogate model. Our approach learns a simulation as a composition of three functions: an encoder, a processor and a decoder. The encoder takes in a graph representation of a 3D geometry of an injection molding part and returns a numeric embedding of each node in the graph. The processor updates the embeddings of each node multiple times based on its neighbors. The decoder then decodes the final embeddings of each node into physically meaningful variables, say, the fill state of the node.
Our model can predict the progression of the flow front during a time step with a fixed size. To simulate a full mold filling process, our model is applied sequentially until the entire mold is filled. Our architecture is applicable to any kind of material, geometry and injection process parameters. We evaluate our architecture by its accuracy and runtime when predicting node properties. We also evaluate our models transfer learning ability on a real world injection molding part.
|
| Vortragende(r)
|
Mingzhe Tao
|
| Titel
|
Meta-learning for Encoder Selection
|
| Vortragstyp
|
Proposal
|
| Betreuer(in)
|
Federico Matteucci
|
| Vortragssprache
|
|
| Vortragsmodus
|
in Präsenz
|
| Kurzfassung
|
In the real world, mixed-type data is commonly used, which means it contains both categorical and numerical data. However, most algorithms can only learn from numerical data. This makes the selection of encoder becoming very important. In this presentation, I will present an approach by using ideas from meta-learning to predict the performance from the meta-features and encoders.
|
- Neuen Vortrag erstellen
Hinweise
