Trends in Computer Vision, NLP, Time Series Analysis and Combinatorial Optimization
Practical information:
Georganiseerd door:
Programme
Trends in Computer Vision
Thursday 16 March 2023 (9 - 12.30h)
Computer Vision traditionally employs supervised Deep Learning of Convolutional Neural Networks. However, to reach adequate performance in industrial application cases, it typically needs an enormous amount of annotated training examples. We present various techniques to tackle this:
- Few Shot Learning and Semi Supervised Learning
- Training from simulation images, Domain Transfer and Generative Adversarial Networks
Next to that, we focus on techniques to drastically reduce the computational demand of CNN-based neural networks, in order to reach real-time behaviour on resource-constrained embedded platforms.
This seminar is taught by Prof. Toon Goedemé.
Use case: Unilin (Beau Boutmans)
Trends in Natural Language Processing
Thursday 16 March 2023 (13 - 16.30h)
The second seminar will focus on the newest techniques in Natural Language Processing, with a particular focus on the BERT and transformer architecture. The following topics will be tackled during the seminar:
- Overview of BERT and transformer architecture
- Transfer learning in NLP
- Integration of CV and NLP
- Explainability
- Multilinguality
This seminar is taught by dr. Katrien Laenen.
Use case: ML6 about building a knowledge sharing platform for a large legal institution (Mathias Leys)
Time Series Analysis
Thursday 23 March 2023 (9 - 12.30h)
The domain of Time Series Analysis has mainly focussed on static data sets and historical data. However, when performing analyses in an online setting, the distribution of data changes over time and adaptive learning strategies are vital. This seminar will teach you about techniques to handle this continuously changing data stream, such as Automated Machine Learning, Adaptive Learning, Continual Learning, Transfer Learning and Active Learning.
This seminar is taught by Prof. Mathias Verbeke.
Use case: Skyline Communications (Dennis Dreesen and Veerle Ledoux)
Combinatorial Optimization
Thursday 23 March 2023 (13 - 16.30h)
The third seminar will give you an overview of the huge progress that has been made regarding calculation speed and efficiency in solving combinatorial problems, before addressing these topics:
- Uncertainty within combinatorial optimization, and dynamic problems
- Data science meets optimization
- Parameter tuning and instance space analysis
- GPU and Cloud based optimization
This seminar is taught by Prof. Tony Wauters.
Use case: ARDIS (Bern De Lille & Koen Handekyn)
