Announcements
Research News
This thesis proposes a transformer-based framework with post-inference strategies designed to improve temporal coherence and semantic consistency in range-based, multiclass anomaly localization for multivariate time series.
Date: 25.06.2025 / 09:00 Place: A-212
The thesis examines reference selection in Turkish, focusing on how linguistic and cognitive factors shape the choice of referring expressions. A novel coreference corpus based on spontaneous, goal-directed dialogs was developed using a new annotation scheme. Statistical and machine learning analyses reveal that competition and distance metrics are the strongest predictors of referential form, while grammatical and speaker roles show weaker but significant effects. The findings support and extend key predictions from theories about reference production and comprehension by focusing on a typologically different language and using naturalistic data.
Date: 18.06.2025 / 10:00 Place: A-212
As next-generation avionic platforms become increasingly connected, systems that were once isolated are now exposed to cyber threats. Communication standards like MIL-STD-1553, widely used in commercial, military, and aerospace platforms, were originally designed without security considerations, resulting in growing vulnerabilities. Implementing conventional security upgrades is costly and brings certification challenges. Intrusion Detection Systems (IDS) offer a non-intrusive alternative, requiring no hardware or software changes. This study aims to enhance the security of MIL-STD-1553 communication buses by integrating a hardware fingerprinting-based IDS and evaluates the effectiveness of machine and deep learning methods in detecting unauthorized devices on the bus.
Date: 02.06.2025 / 11:00 Place: A-212
This thesis aims to reduce the need for pixel-level labeled data for semantic segmentation. A DeepLabV3+ based model trained on synthetic images is supplemented with a Domain-Adversarial Neural Network (DANN), an adversarial domain adaptation method, to adapt to real images. The model is applied in unsupervised and semi-supervised domain adaptation scenarios. In the semi-supervised adaptation method in particular, similar performance was achieved using 92% less labeled real data compared to the DeepLabV3+ method trained with supervised learning and without domain adaptation. This study provides an effective solution that reduces the burden of image labeling.
Date: 26.05.2025 / 15:00 Place: A-212
A common evolutionary origin of the Obscurin and Trio protein families was identified through phylogenetic analyses, indicating ancient domain shuffling events across vertebrate lineages. Building on these findings, the novel variant effect predictor TrioNsight was developed for TrioN-like DH domains. The predictor, which integrates structural, evolutionary and physicochemical features, achieved high accuracy in distinguishing pathogenic variants (MCC: 0.906; 1.0 indicates perfect prediction). This work bridges fundamental evolutionary biology and clinical application, offering both mechanistic insights into protein family evolution and a practical tool for the interpretation of disease-associated variants in functionally critical domains.
Date: 30.05.2025 / 10:00 Place: A-212
