Yazılım Mühendisliği
Permanent URI for this community
Browse
Browsing Yazılım Mühendisliği by Issue Date
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Transfer learning for continuous control(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2019., 2019.) Ada, Suzan Ece.; Akın, H. Levent.Agentstrainedwithdeepreinforcementlearningalgorithmsarecapableofperforming highly complex tasks including locomotion in continuous environments. In order to attain a human-level performance, the next step of research should be to investigate the ability to transfer the learning acquired in one task to unknown tasks. Concerns on generalization and overfitting in deep reinforcement learning are not usually addressed in current transfer learning research. This issue results in simplistic benchmarks and inaccurate algorithm comparisons due to rudimentary assessments. In this thesis, we propose novel regularization techniques exclusive to policy gradient algorithms for continuous control through the application of sample elimination and early stopping. By discarding samples that lead to overfitting via strict clipping we will generate robust policies for a humanoid with high generalization capacity. We also suggest the inclusion of training iteration to the hyperparameters in deep transfer learning problems. We recommend resorting to earlier snapshots of parameters depending on the target task due to the occurrence of overfitting to the source task. We demonstrate that a humanoid is capable of performing forward locomotion in unseen environments with different gravities and tangential frictions using strict clipping and early stopping. Furthermore, we evaluate our propositions on a delivery task where a humanoid is required to carry a heavy box while walking and inter-robot transfer tasks where the humanoid transfers its learning to taller and shorter robots. Because source task performance is not indicative of the generalization capacity of the algorithm we propose three different transfer learning evaluation methods. We increase the generalization capacity of a state-of-art adversarial algorithm by introducing entropy bonus, proposing different critic architectures and using simpler adversaries. Finally, we evaluate the robustness of these adversarial algorithms on morphologically modified hopper environments and environments with unknown gravities according to the criteria we proposed.Item Improving spatial reuse and throughput via OBSS/PD in IEEE 802.11ax(Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2021., 2021.) Karakoç, Ali.; Yılmaz, H. Birkan.; Kuran, Mehmet Şükrü.The aim of this thesis is to improve spatial reuse by using special mechanisms in IEEE 802.11ax wireless local area networks. If there are numbers of basic service sets (BSSs) in the same vicinity, overlapping BSSs may create interference. In this situation, BSSs cannot transmit simultaneously, because of the collisions. Moreover, there may be spectral e ciency problems like \Hidden Node Problem" or \Exposed Node Problem", that may cause signi cant performance degradation to the system. In IEEE 802.11ax amendment, to address these spectral e ciency problems, a couple of mechanisms are introduced. One of the spectral e ciency mechanisms that address these problems is the overlapping BSS preamble detection (OBSS/PD) mechanism. OBSS/PD is a color-based mechanism that is used to detect and prevent overlapping BSS interference e ciently. In this thesis, we propose a rate-adaptive dynamic OBSS/PD threshold algorithm that dynamically adjusts the OBSS/PD threshold with respect to the changes in the channel conditions and selected data rates. Additionally, hidden and exposed node problems due to overlapping BSS are reduced. The proposed mechanism is designed to work along with the rate selection algorithms. In this study, the scenarios are performed with Minstrel and Thompson rate selection algorithms. The performance of the proposed mechanism has been compared with the legacy carrier sensitivity threshold algorithms (DSC and RTOT). Our algorithm has shown more stable performance than the reference algorithms in the Minstrel scenario, and none of the threshold algorithms have shown a signi cant performance enhancement relative to the others. When the Thompson rate selection algorithm is used, our proposed algorithm has shown a better performance and stability than the legacy carrier sensitivity threshold algorithms.