| 1 |
| Title |
|
Unleashing the Power of Data: Techniques for Effective Feature Engineering |
| Speaker |
|
Tanzeela Javid Kaloo |
| Advisors |
|
Ender Mete Ekşioğlu |
| Date |
|
May 17, 2023 |
| Time |
|
15:30 (GMT +3) |
Abstract
The process of feature extraction involves picking out important properties from data while ignoring unnecessary or unimportant ones. Good feature engineering is crucial because it can highlight trends and connections that could be obscured in raw data. It is a crucial stage in creating machine learning models since it has a big effect on how well they operate. Statistics, subject area knowledge, and creative thinking are all essential skills for feature engineering. Feature extraction, selection, and treatment of missing data, outliers, and unbalanced classes are all part of the procedure. By incorporating both fresh and existing features, feature extraction algorithms are able to recognize complicated relationships. To make sure that characteristics are comparable and the same size, normalizing and scaling are often employed procedures. Moreover, categorical variables can be binned and encoded to transform features into numerical data that machine learning models can use. An emerging field called automated feature engineering uses machine learning to find and create the best features for improved model performance. |
| • |
| 2 |
| Title |
|
Investigation of Drug Resistance Mechanisms for Antiandrogen Prostate Cancer Drug Enzalutamide using Molecular Dynamics Simulations |
| Speaker |
|
Behzad Aslani Avilaq |
| Advisors |
|
Sefer Baday |
| Date |
|
May 17, 2023 |
| Time |
|
16:00 (GMT +3) |
Abstract
Prostate cancer is the second most common cancer type in men after lung cancer. Androgen Receptor (AR) signaling is responsible for cell growth and duplication in prostatic cells. In time of this study, several drug molecules have been developed for treatment of prostate cancer. In clinical studies, Enzalutamide (ENZ) as a second-generation anti-androgenic drug is shown to reduce radiographic progression significantly and delay the need for chemotherapy. ENZ inhibits the AR and prevents the AR signaling; however, after some time resistance against ENZ occurs, and two-point mutations as F876I, and F876L abolish the antagonistic activity of ENZ. Literature suggests that displacement or instability of Helix12 (H12) might be causing the switch between antagonistic to agonistic activity of AR. However, specific interactions that lead to instability of helix12 remain unclear. In the current study we performed μs-long AR-ENZ Molecular Dynamics (MD) simulations and utilized advanced analysis techniques. Results suggest that upon ENZ binding, unbending of H11 leads to loss in stability of H12. In other words, this study found an important pathway of interactions in AR that change the antagonist-AR to agonist- AR; thus, resulting in resistance against ENZ. |
| • |
| 3 |
| Title |
|
User Authentication through Mouse Dynamics Paper Replication |
| Speaker |
|
Oğuzhan Salman |
| Advisors |
|
Kemal Bicakci |
| Date |
|
May 24, 2023 |
| Time |
|
15:30 (GMT +3) |
Abstract
The field of Mouse Dynamics seeks to mitigate the risk of unauthorized access to computer systems through the utilization of behavioral analytics, which entails the examination of an individual’s mouse usage during digital interactions with the aim of identification. Despite the limited literature pertaining to the refinement of the accuracy of Mouse Dynamics, as of 2022, there are currently no commercially available products that allow for the practical application of this technology in daily life. While analogous techniques, such as Keyboard Dynamics, which employ keystroke analysis, have been implemented in commercial products, Mouse Dynamics remains unavailable to the general population. Therefore, the objective of this paper is to replicate the findings of a recent study and investigate the feasibility of implementing Mouse Dynamics for the general populace. |
| • |
| 4 |
| Title |
|
Energy Efficient Offloading Decision for Multi-Access Edge Computing Enabled UAV Swarms |
| Speaker |
|
Homa Maleki |
| Advisors |
|
Lutfiye Durak-Ata |
| Date |
|
May 24, 2023 |
| Time |
|
16:00 (GMT +3) |
Abstract
Recent developments in unmanned aerial vehicle technology have given UAVs more processing and storage resources, paving the way for the concept of edge computing-enabled UAV networks. In this project, we investigate a cooperative multi-agent computation offloading framework in a UAV swarm in which flying UAVs with missions can offload part of the missions to neighboring UAVs or fixed edge servers placed at Next Generation Node B (gNB) to decrease the total energy consumption of all devices during a core mission by leading them to make steady sequences of offloading decisions under uncertainties of the dynamic environment. |