Barış Akgün

Raphael Trumpp, Ömer Veysel Çağatan, Barış Akgün et al.

Pixel-based deep reinforcement learning agents are typically trained on heavily downsampled visual observations, a convention inherited from early benchmarks rather than grounded in principled design. In this work, we show that observation resolution is a critical yet overlooked variable for policy learning: higher-resolution inputs can substantially improve both performance and generalization, provided the network architecture can process them effectively. We find that the widely used Impala encoder, which flattens spatial features into a vector, suffers from quadratic parameter growth as resolution increases and fails to leverage the additional visual detail. Replacing this operation with global average pooling, as in the Impoola architecture, decouples parameter count from resolution and yields consistent improvements across resolutions and network widths - at their respective best conditions, visual scaling unlocks a 28 % performance gain for Impoola over Impala. These gains are strongest in environments that require precise perception of small or distant objects, and gradient saliency analysis confirms that the underlying mechanism is a more spatially localized visual attention of the policy at higher resolutions. Our results challenge the prevailing practice of aggressive input downsampling and position resolution-independent architectures as a simple, effective path toward scalable visual deep RL. To facilitate future research on resolution scaling in deep RL, we publicly release the open-source code for the Procgen-HD benchmark: https://github.com/raphajaner/procgen-hd.

Ahnaf Hannan Lodhi, Barış Akgün, Öznur Özkasap

Federated Learning (FL) has become a key choice for distributed machine learning. Initially focused on centralized aggregation, recent works in FL have emphasized greater decentralization to adapt to the highly heterogeneous network edge. Among these, Hierarchical, Device-to-Device and Gossip Federated Learning (HFL, D2DFL \& GFL respectively) can be considered as foundational FL algorithms employing fundamental aggregation strategies. A number of FL algorithms were subsequently proposed employing multiple fundamental aggregation schemes jointly. Existing research, however, subjects the FL algorithms to varied conditions and gauges the performance of these algorithms mainly against Federated Averaging (FedAvg) only. This work consolidates the FL landscape and offers an objective analysis of the major FL algorithms through a comprehensive cross-evaluation for a wide range of operating conditions. In addition to the three foundational FL algorithms, this work also analyzes six derived algorithms. To enable a uniform assessment, a multi-FL framework named FLAGS: Federated Learning AlGorithms Simulation has been developed for rapid configuration of multiple FL algorithms. Our experiments indicate that fully decentralized FL algorithms achieve comparable accuracy under multiple operating conditions, including asynchronous aggregation and the presence of stragglers. Furthermore, decentralized FL can also operate in noisy environments and with a comparably higher local update rate. However, the impact of extremely skewed data distributions on decentralized FL is much more adverse than on centralized variants. The results indicate that it may not be necessary to restrict the devices to a single FL algorithm; rather, multi-FL nodes may operate with greater efficiency.

Bugra C. Sefercik, Baris Akgun

Traditional approaches to extrinsic calibration use fiducial markers and learning-based approaches rely heavily on simulation data. In this work, we present a learning-based markerless extrinsic calibration system that uses a depth camera and does not rely on simulation data. We learn models for end-effector (EE) segmentation, single-frame rotation prediction and keypoint detection, from automatically generated real-world data. We use a transformation trick to get EE pose estimates from rotation predictions and a matching algorithm to get EE pose estimates from keypoint predictions. We further utilize the iterative closest point algorithm, multiple-frames, filtering and outlier detection to increase calibration robustness. Our evaluations with training data from multiple camera poses and test data from previously unseen poses give sub-centimeter and sub-deciradian average calibration and pose estimation errors. We also show that a carefully selected single training pose gives comparable results.

Ege Onat Özsüer, Barış Akgün, Fatma Güney

Reinforcement Learning (RL) has the potential to surpass human performance in driving without needing any expert supervision. Despite its promise, the state-of-the-art in sensorimotor self-driving is dominated by imitation learning methods due to the inherent shortcomings of RL algorithms. Nonetheless, RL agents are able to discover highly successful policies when provided with privileged ground truth representations of the environment. In this work, we investigate what separates privileged RL agents from sensorimotor agents for urban driving in order to bridge the gap between the two. We propose vision-based deep learning models to approximate the privileged representations from sensor data. In particular, we identify aspects of state representation that are crucial for the success of the RL agent such as desired route generation and stop zone prediction, and propose solutions to gradually develop less privileged RL agents. We also observe that bird's-eye-view models trained on offline datasets do not generalize to online RL training due to distribution mismatch. Through rigorous evaluation on the CARLA simulation environment, we shed light on the significance of the state representations in RL for autonomous driving and point to unresolved challenges for future research.

Farzin Negahbani, Onur Berk Töre, Fatma Güney et al.

Most autonomous vehicles are equipped with LiDAR sensors and stereo cameras. The former is very accurate but generates sparse data, whereas the latter is dense, has rich texture and color information but difficult to extract robust 3D representations from. In this paper, we propose a novel data fusion algorithm to combine accurate point clouds with dense but less accurate point clouds obtained from stereo pairs. We develop a framework to integrate this algorithm into various 3D object detection methods. Our framework starts with 2D detections from both of the RGB images, calculates frustums and their intersection, creates Pseudo-LiDAR data from the stereo images, and fills in the parts of the intersection region where the LiDAR data is lacking with the dense Pseudo-LiDAR points. We train multiple 3D object detection methods and show that our fusion strategy consistently improves the performance of detectors.

Ahnaf Hannan Lodhi, Barış Akgün, Öznur Özkasap

The potential held by the gargantuan volumes of data being generated across networks worldwide has been truly unlocked by machine learning techniques and more recently Deep Learning. The advantages offered by the latter have seen it rapidly becoming a framework of choice for various applications. However, the centralization of computational resources and the need for data aggregation have long been limiting factors in the democratization of Deep Learning applications. Edge Computing is an emerging paradigm that aims to utilize the hitherto untapped processing resources available at the network periphery. Edge Intelligence (EI) has quickly emerged as a powerful alternative to enable learning using the concepts of Edge Computing. Deep Learning-based Edge Intelligence or Deep Edge Intelligence (DEI) lies in this rapidly evolving domain. In this article, we provide an overview of the major constraints in operationalizing DEI. The major research avenues in DEI have been consolidated under Federated Learning, Distributed Computation, Compression Schemes and Conditional Computation. We also present some of the prevalent challenges and highlight prospective research avenues.

Deniz Lefkeli, Baris Akgun, Sahibzada Omar et al.

Recent research on human robot interaction explored whether people's tendency to conform to others extends to artificial agents (Hertz & Wiese, 2016). However, little is known about to what extent perception of a robot as having a mind affects people's decisions. Grounded on the theory of mind perception, the current study proposes that artificial agents can induce decision change to the extent in which individuals perceive them as having minds. By varying the degree to which robots expressed ability to act (agency) or feel (experience), we specifically investigated the underlying mechanisms of mind attribution to robots and social influence. Our results show an interactive effect of perceived experience and perceived agency on social influence induced by artificial agents. The findings provide preliminary insights regarding autonomous robots' influence on individuals' decisions and form a basis for understanding the underlying dynamics of decision making with robots.

Cem Eteke, Hayati Havlucu, Nisa İrem Kırbaç et al.

Wearable devices have the potential to enhance sports performance, yet they are not fulfilling this promise. Our previous studies with 6 professional tennis coaches and 20 players indicate that this could be due the lack of psychological or mental state feedback, which the coaches claim to provide. Towards this end, we propose to detect the flow state, mental state of optimal performance, using wearables data to be later used in training. We performed a study with a professional tennis coach and two players. The coach provided labels about the players' flow state while each player had a wearable device on their racket holding wrist. We trained multiple models using the wearables data and the coach labels. Our deep neural network models achieved around 98% testing accuracy for a variety of conditions. This suggests that the flow state or what coaches recognize as flow, can be detected using wearables data in tennis which is a novel result. The implication for the HCI community is that having access to such information would allow for design of novel hardware and interaction paradigms that would be helpful in professional athlete training.