Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
aira:start [2026/11/12 08:59] – [Schedule Autumn 2027] mzkaira:start [2026/11/12 09:06] (current) – [2025-11-12] mzk
Line 608: Line 608:
  
  
 +
 +==== 2026-11-12 ====
 +<WRAP column 15%>
 +{{ :aira:damian-kucharski-foto.png?200| }}
 +</WRAP>
 +
 +<WRAP column 75%>
 +
 +**Speaker**: Damian Kucharski,  PhD Candidate @ Silesian University of Technology
 +
 +**Title**: Automated Ensemble Design in the Wild.
 +
 +**Abstract**:
 +Combining machine learning models can improve predictive performance, but deciding which models to include and how to combine their predictions remains a challenging design problem. More flexible ensemble structures offer additional opportunities to exploit complementary errors, while also increasing complexity and the risk of overfitting the validation data used to guide their construction.
 +This talk presents OKAPI, a multi-objective genetic programming method that jointly searches for ensemble membership and prediction fusion procedures. Ensembles are represented as stateful trees that accumulate and transform model predictions through an evolved sequence of operations. Predictive performance and structural complexity are treated as separate objectives, allowing the search to discover compact ensembles with inspectable fusion procedures.
 +Experiments spanning medical image classification, methane plume segmentation in satellite imagery, and tabular classification examine both the capabilities and limitations of this approach. Particular attention is given to the relationship between validation improvements and performance on unseen data, the role of complexity control, and the distinction between individual model quality and contribution within an ensemble. The talk also explores prior anchoring, which combines an evolved fusion procedure with a fixed averaging reference to limit overfitting. Together, these findings highlight the conditions under which automated ensemble design can translate search flexibility into useful predictive performance.
 +
 +**Biogram**: 
 +Damian Kucharski received his BEng (2021) and MSc (2022) degrees in Computer Science from the Silesian University of Technology, Poland, where he recently submitted his doctoral thesis on automated ensemble design. His research focuses on evolutionary computation and the automated design and optimisation of machine learning ensembles, with applications in biomedical data analysis and Earth observation. Alongside his academic work, he has several years of industrial experience in machine learning and deep learning, including medical imaging, processing large datasets, and model optimisation. His interests also include GPU computing and efficient model training and inference.
 +</WRAP>
 +<WRAP clear></WRAP>
  
 ==== 2026-10-29 ==== ==== 2026-10-29 ====
aira/start.1794473991.txt.gz · Last modified: 2026/11/12 08:59 by mzk
Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0