Senem Velipasalar

Chenbin Pan, Rui Hou, Hanchao Yu et al. · meta-ai

Whether by processing videos with fixed resolution from start to end or incorporating pooling and down-scaling strategies, existing video transformers process the whole video content throughout the network without specially handling the large portions of redundant information. In this paper, we present a Supertoken Video Transformer (SVT) that incorporates a Semantic Pooling Module (SPM) to aggregate latent representations along the depth of visual transformer based on their semantics, and thus, reduces redundancy inherent in video inputs.~Qualitative results show that our method can effectively reduce redundancy by merging latent representations with similar semantics and thus increase the proportion of salient information for downstream tasks.~Quantitatively, our method improves the performance of both ViT and MViT while requiring significantly less computations on the Kinectics and Something-Something-V2 benchmarks.~More specifically, with our SPM, we improve the accuracy of MAE-pretrained ViT-B and ViT-L by 1.5% with 33% less GFLOPs and by 0.2% with 55% less FLOPs, respectively, on the Kinectics-400 benchmark, and improve the accuracy of MViTv2-B by 0.2% and 0.3% with 22% less GFLOPs on Kinectics-400 and Something-Something-V2, respectively.

Md Zahid Hasan, Jiajing Chen, Jiyang Wang et al.

Recognizing the activities causing distraction in real-world driving scenarios is critical for ensuring the safety and reliability of both drivers and pedestrians on the roadways. Conventional computer vision techniques are typically data-intensive and require a large volume of annotated training data to detect and classify various distracted driving behaviors, thereby limiting their efficiency and scalability. We aim to develop a generalized framework that showcases robust performance with access to limited or no annotated training data. Recently, vision-language models have offered large-scale visual-textual pretraining that can be adapted to task-specific learning like distracted driving activity recognition. Vision-language pretraining models, such as CLIP, have shown significant promise in learning natural language-guided visual representations. This paper proposes a CLIP-based driver activity recognition approach that identifies driver distraction from naturalistic driving images and videos. CLIP's vision embedding offers zero-shot transfer and task-based finetuning, which can classify distracted activities from driving video data. Our results show that this framework offers state-of-the-art performance on zero-shot transfer and video-based CLIP for predicting the driver's state on two public datasets. We propose both frame-based and video-based frameworks developed on top of the CLIP's visual representation for distracted driving detection and classification tasks and report the results.

Geethu Joseph, Chen Zhong, M. Cenk Gursoy et al.

In this paper, we address the problem of detecting anomalies among a given set of binary processes via learning-based controlled sensing. Each process is parameterized by a binary random variable indicating whether the process is anomalous. To identify the anomalies, the decision-making agent is allowed to observe a subset of the processes at each time instant. Also, probing each process has an associated cost. Our objective is to design a sequential selection policy that dynamically determines which processes to observe at each time with the goal to minimize the delay in making the decision and the total sensing cost. We cast this problem as a sequential hypothesis testing problem within the framework of Markov decision processes. This formulation utilizes both a Bayesian log-likelihood ratio-based reward and an entropy-based reward. The problem is then solved using two approaches: 1) a deep reinforcement learning-based approach where we design both deep Q-learning and policy gradient actor-critic algorithms; and 2) a deep active inference-based approach. Using numerical experiments, we demonstrate the efficacy of our algorithms and show that our algorithms adapt to any unknown statistical dependence pattern of the processes.

Feng Wang, M. Cenk Gursoy, Senem Velipasalar

Federated learning has attracted growing interest as it preserves the clients' privacy. As a variant of federated learning, federated transfer learning utilizes the knowledge from similar tasks and thus has also been intensively studied. However, due to the limited radio spectrum, the communication efficiency of federated learning via wireless links is critical since some tasks may require thousands of Terabytes of uplink payload. In order to improve the communication efficiency, we in this paper propose the feature-based federated transfer learning as an innovative approach to reduce the uplink payload by more than five orders of magnitude compared to that of existing approaches. We first introduce the system design in which the extracted features and outputs are uploaded instead of parameter updates, and then determine the required payload with this approach and provide comparisons with the existing approaches. Subsequently, we analyze the random shuffling scheme that preserves the clients' privacy. Finally, we evaluate the performance of the proposed learning scheme via experiments on an image classification task to show its effectiveness.

Feng Wang, Senem Velipasalar, M. Cenk Gursoy

Federated learning (FL) aims at keeping client data local to preserve privacy. Instead of gathering the data itself, the server only collects aggregated gradient updates from clients. Following the popularity of FL, there has been considerable amount of work, revealing the vulnerability of FL approaches by reconstructing the input data from gradient updates. Yet, most existing works assume an FL setting with unrealistically small batch size, and have poor image quality when the batch size is large. Other works modify the neural network architectures or parameters to the point of being suspicious, and thus, can be detected by clients. Moreover, most of them can only reconstruct one sample input from a large batch. To address these limitations, we propose a novel and completely analytical approach, referred to as the maximum knowledge orthogonality reconstruction (MKOR), to reconstruct clients' input data. Our proposed method reconstructs a mathematically proven high quality image from large batches. MKOR only requires the server to send secretly modified parameters to clients and can efficiently and inconspicuously reconstruct the input images from clients' gradient updates. We evaluate MKOR's performance on the MNIST, CIFAR-100, and ImageNet dataset and compare it with the state-of-the-art works. The results show that MKOR outperforms the existing approaches, and draws attention to a pressing need for further research on the privacy protection of FL so that comprehensive defense approaches can be developed.

Feng Wang, M. Cenk Gursoy, Senem Velipasalar

In this paper, we present a multi-agent deep reinforcement learning (deep RL) framework for network slicing in a dynamic environment with multiple base stations and multiple users. In particular, we propose a novel deep RL framework with multiple actors and centralized critic (MACC) in which actors are implemented as pointer networks to fit the varying dimension of input. We evaluate the performance of the proposed deep RL algorithm via simulations to demonstrate its effectiveness. Subsequently, we develop a deep RL based jammer with limited prior information and limited power budget. The goal of the jammer is to minimize the transmission rates achieved with network slicing and thus degrade the network slicing agents' performance. We design a jammer with both listening and jamming phases and address jamming location optimization as well as jamming channel optimization via deep RL. We evaluate the jammer at the optimized location, generating interference attacks in the optimized set of channels by switching between the jamming phase and listening phase. We show that the proposed jammer can significantly reduce the victims' performance without direct feedback or prior knowledge on the network slicing policies. Finally, we devise a Nash-equilibrium-supervised policy ensemble mixed strategy profile for network slicing (as a defensive measure) and jamming. We evaluate the performance of the proposed policy ensemble algorithm by applying on the network slicing agents and the jammer agent in simulations to show its effectiveness.

Chenbin Pan, Zhiqi Zhang, Senem Velipasalar et al.

Capturing interaction of hands with objects is important to autonomously detect human actions from egocentric videos. In this work, we present a pyramid video transformer with a dynamic class token generator for egocentric action recognition. Different from previous video transformers, which use the same static embedding as the class token for diverse inputs, we propose a dynamic class token generator that produces a class token for each input video by analyzing the hand-object interaction and the related motion information. The dynamic class token can diffuse such information to the entire model by communicating with other informative tokens in the subsequent transformer layers. With the dynamic class token, dissimilarity between videos can be more prominent, which helps the model distinguish various inputs. In addition, traditional video transformers explore temporal features globally, which requires large amounts of computation. However, egocentric videos often have a large amount of background scene transition, which causes discontinuities across distant frames. In this case, blindly reducing the temporal sampling rate will risk losing crucial information. Hence, we also propose a pyramid architecture to hierarchically process the video from short-term high rate to long-term low rate. With the proposed architecture, we significantly reduce the computational cost as well as the memory requirement without sacrificing from the model performance. We perform comparisons with different baseline video transformers on the EPIC-KITCHENS-100 and EGTEA Gaze+ datasets. Both quantitative and qualitative results show that the proposed model can efficiently improve the performance for egocentric action recognition.

Chenbin Pan, Burhaneddin Yaman, Tommaso Nesti et al.

Autonomous driving is a complex and challenging task that aims at safe motion planning through scene understanding and reasoning. While vision-only autonomous driving methods have recently achieved notable performance, through enhanced scene understanding, several key issues, including lack of reasoning, low generalization performance and long-tail scenarios, still need to be addressed. In this paper, we present VLP, a novel Vision-Language-Planning framework that exploits language models to bridge the gap between linguistic understanding and autonomous driving. VLP enhances autonomous driving systems by strengthening both the source memory foundation and the self-driving car's contextual understanding. VLP achieves state-of-the-art end-to-end planning performance on the challenging NuScenes dataset by achieving 35.9\% and 60.5\% reduction in terms of average L2 error and collision rates, respectively, compared to the previous best method. Moreover, VLP shows improved performance in challenging long-tail scenarios and strong generalization capabilities when faced with new urban environments.

Minmin Yang, Huantao Ren, Senem Velipasalar

Existing zero-shot 3D point cloud segmentation methods often struggle with limited transferability from seen classes to unseen classes and from semantic to visual space. To alleviate this, we introduce 3D-PointZshotS, a geometry-aware zero-shot segmentation framework that enhances both feature generation and alignment using latent geometric prototypes (LGPs). Specifically, we integrate LGPs into a generator via a cross-attention mechanism, enriching semantic features with fine-grained geometric details. To further enhance stability and generalization, we introduce a self-consistency loss, which enforces feature robustness against point-wise perturbations. Additionally, we re-represent visual and semantic features in a shared space, bridging the semantic-visual gap and facilitating knowledge transfer to unseen classes. Experiments on three real-world datasets, namely ScanNet, SemanticKITTI, and S3DIS, demonstrate that our method achieves superior performance over four baselines in terms of harmonic mIoU. The code is available at \href{https://github.com/LexieYang/3D-PointZshotS}{Github}.

Yantao Lu, Xuetao Hao, Yilan Li et al.

3D object detection from LiDAR data for autonomous driving has been making remarkable strides in recent years. Among the state-of-the-art methodologies, encoding point clouds into a bird's eye view (BEV) has been demonstrated to be both effective and efficient. Different from perspective views, BEV preserves rich spatial and distance information between objects. Yet, while farther objects of the same type do not appear smaller in the BEV, they contain sparser point cloud features. This fact weakens BEV feature extraction using shared-weight convolutional neural networks (CNNs). In order to address this challenge, we propose Range-Aware Attention Network (RAANet), which extracts effective BEV features and generates superior 3D object detection outputs. The range-aware attention (RAA) convolutions significantly improve feature extraction for near as well as far objects. Moreover, we propose a novel auxiliary loss for point density estimation to further enhance the detection accuracy of RAANet for occluded objects. It is worth to note that our proposed RAA convolution is lightweight and compatible to be integrated into any CNN architecture used for detection from a BEV. Extensive experiments on the nuScenes and KITTI datasets demonstrate that our proposed approach outperforms the state-of-the-art methods for LiDAR-based 3D object detection, with real-time inference speed of 16 Hz for the full version and 22 Hz for the lite version tested on nuScenes lidar frames. The code is publicly available at our Github repository https://github.com/erbloo/RAAN.

Yantao Lu, Yunhan Jia, Jianyu Wang et al.

Neural networks are known to be vulnerable to carefully crafted adversarial examples, and these malicious samples often transfer, i.e., they remain adversarial even against other models. Although great efforts have been delved into the transferability across models, surprisingly, less attention has been paid to the cross-task transferability, which represents the real-world cybercriminal's situation, where an ensemble of different defense/detection mechanisms need to be evaded all at once. In this paper, we investigate the transferability of adversarial examples across a wide range of real-world computer vision tasks, including image classification, object detection, semantic segmentation, explicit content detection, and text detection. Our proposed attack minimizes the ``dispersion'' of the internal feature map, which overcomes existing attacks' limitation of requiring task-specific loss functions and/or probing a target model. We conduct evaluation on open source detection and segmentation models as well as four different computer vision tasks provided by Google Cloud Vision (GCV) APIs, to show how our approach outperforms existing attacks by degrading performance of multiple CV tasks by a large margin with only modest perturbations linf=16.

Chenbin Pan, Burhaneddin Yaman, Senem Velipasalar et al.

Autonomous driving stands as a pivotal domain in computer vision, shaping the future of transportation. Within this paradigm, the backbone of the system plays a crucial role in interpreting the complex environment. However, a notable challenge has been the loss of clear supervision when it comes to Bird's Eye View elements. To address this limitation, we introduce CLIP-BEVFormer, a novel approach that leverages the power of contrastive learning techniques to enhance the multi-view image-derived BEV backbones with ground truth information flow. We conduct extensive experiments on the challenging nuScenes dataset and showcase significant and consistent improvements over the SOTA. Specifically, CLIP-BEVFormer achieves an impressive 8.5\% and 9.2\% enhancement in terms of NDS and mAP, respectively, over the previous best BEV model on the 3D object detection task.

Feng Wang, M. Cenk Gursoy, Senem Velipasalar

In this paper, we propose feature-based federated transfer learning as a novel approach to improve communication efficiency by reducing the uplink payload by multiple orders of magnitude compared to that of existing approaches in federated learning and federated transfer learning. Specifically, in the proposed feature-based federated learning, we design the extracted features and outputs to be uploaded instead of parameter updates. For this distributed learning model, we determine the required payload and provide comparisons with the existing schemes. Subsequently, we analyze the robustness of feature-based federated transfer learning against packet loss, data insufficiency, and quantization. Finally, we address privacy considerations by defining and analyzing label privacy leakage and feature privacy leakage, and investigating mitigating approaches. For all aforementioned analyses, we evaluate the performance of the proposed learning scheme via experiments on an image classification task and a natural language processing task to demonstrate its effectiveness.

Huantao Ren, Jiajing Chen, Senem Velipasalar

Gait is a behavioral biometric modality that can be used to recognize individuals by the way they walk from a far distance. Most existing gait recognition approaches rely on either silhouettes or skeletons, while their joint use is underexplored. Features from silhouettes and skeletons can provide complementary information for more robust recognition against appearance changes or pose estimation errors. To exploit the benefits of both silhouette and skeleton features, we propose a new gait recognition network, referred to as the GaitPoint+. Our approach models skeleton key points as a 3D point cloud, and employs a computational complexity-conscious 3D point processing approach to extract skeleton features, which are then combined with silhouette features for improved accuracy. Since silhouette- or CNN-based methods already require considerable amount of computational resources, it is preferable that the key point learning module is faster and more lightweight. We present a detailed analysis of the utilization of every human key point after the use of traditional max-pooling, and show that while elbow and ankle points are used most commonly, many useful points are discarded by max-pooling. Thus, we present a method to recycle some of the discarded points by a Recycling Max-Pooling module, during processing of skeleton point clouds, and achieve further performance improvement. We provide a comprehensive set of experimental results showing that (i) incorporating skeleton features obtained by a point-based 3D point cloud processing approach boosts the performance of three different state-of-the-art silhouette- and CNN-based baselines; (ii) recycling the discarded points increases the accuracy further. Ablation studies are also provided to show the effectiveness and contribution of different components of our approach.

Huantao Ren, Minmin Yang, Senem Velipasalar

Deep neural networks have achieved significant success in 3D point cloud classification while relying on large-scale, annotated point cloud datasets, which are labor-intensive to build. Compared to capturing data with LiDAR sensors and then performing annotation, it is relatively easier to sample point clouds from CAD models. Yet, data sampled from CAD models is regular, and does not suffer from occlusion and missing points, which are very common for LiDAR data, creating a large domain shift. Therefore, it is critical to develop methods that can generalize well across different point cloud domains. %In this paper, we focus on the 3D point cloud domain generalization problem. Existing 3D domain generalization methods employ point-based backbones to extract point cloud features. Yet, by analyzing point utilization of point-based methods and observing the geometry of point clouds from different domains, we have found that a large number of point features are discarded by point-based methods through the max-pooling operation. This is a significant waste especially considering the fact that domain generalization is more challenging than supervised learning, and point clouds are already affected by missing points and occlusion to begin with. To address these issues, we propose a novel method for 3D point cloud domain generalization, which can generalize to unseen domains of point clouds. Our proposed method employs multiple 2D projections of a 3D point cloud to alleviate the issue of missing points and involves a simple yet effective convolution-based model to extract features. The experiments, performed on the PointDA-10 and Sim-to-Real benchmarks, demonstrate the effectiveness of our proposed method, which outperforms different baselines, and can transfer well from synthetic domain to real-world domain.

Jiyang Wang, Ayse Altay, Senem Velipasalar

Functional near-infrared spectroscopy (fNIRS) is a non-intrusive way to measure cortical hemodynamic activity. Predicting cognitive workload from fNIRS data has taken on a diffuse set of methods. To be applicable in real-world settings, models are needed, which can perform well across different sessions as well as different subjects. However, most existing works assume that training and testing data come from the same subjects and/or cannot generalize well across never-before-seen subjects. Additional challenges imposed by fNIRS data include the high variations in inter-subject fNIRS data and also in intra-subject data collected across different blocks of sessions. To address these issues, we propose an effective method, referred to as the class-aware-block-aware domain adaptation (CABA-DA) which explicitly minimize intra-session variance by viewing different blocks from the same subject same session as different domains. We minimize the intra-class domain discrepancy and maximize the inter-class domain discrepancy accordingly. In addition, we propose an MLPMixer-based model for cognitive load classification. Experimental results demonstrate the proposed model has better performance compared with three different baseline models on three public-available datasets of cognitive workload. Two of them are collected from n-back tasks and one of them is from finger tapping. From our experiments, we also show the proposed contrastive learning method can also improve baseline models we compared with.

Arda Kabadayi, Senem Velipasalar, Jiajing Chen

Compared to traditional image retrieval tasks, product retrieval in retail settings is even more challenging. Products of the same type from different brands may have highly similar visual appearances, and the query image may be taken from an angle that differs significantly from view angles of the stored catalog images. Foundational models, such as CLIP and SigLIP, often struggle to distinguish these subtle but important local differences. Pixel-wise matching methods, on the other hand, are computationally expensive and incur prohibitively high matching times. In this paper, we propose a new, hybrid method, called PRISM, for product retrieval in retail settings by leveraging the advantages of both vision-language model-based and pixel-wise matching approaches. To provide both efficiency/speed and finegrained retrieval accuracy, PRISM consists of three stages: 1) A vision-language model (SigLIP) is employed first to retrieve the top 35 most semantically similar products from a fixed gallery, thereby narrowing the search space significantly; 2) a segmentation model (YOLO-E) is applied to eliminate background clutter; 3) fine-grained pixel-level matching is performed using LightGlue across the filtered candidates. This framework enables more accurate discrimination between products with high inter-class similarity by focusing on subtle visual cues often missed by global models. Experiments performed on the ABV dataset show that our proposed PRISM outperforms the state-of-the-art image retrieval methods by 4.21% in top-1 accuracy while still remaining within the bounds of real-time processing for practical retail deployments.

Minmin Yang, Huantao Ren, Senem Velipasalar

Cone-beam computed tomography (CBCT) using only a few X-ray projection views enables faster scans with lower radiation dose, but the resulting severe under-sampling causes strong artifacts and poor spatial coverage. We address these challenges in a unified framework. First, we replace conventional UNet/ResNet encoders with TransUNet, a hybrid CNN-Transformer model. Convolutional layers capture local details, while self-attention layers enhance global context. We adapt TransUNet to CBCT by combining multi-scale features, querying view-specific features per 3D point, and adding a lightweight attenuation-prediction head. This yields Trans-CBCT, which surpasses prior baselines by 1.17 dB PSNR and 0.0163 SSIM on the LUNA16 dataset with six views. Second, we introduce a neighbor-aware Point Transformer to enforce volumetric coherence. This module uses 3D positional encoding and attention over k-nearest neighbors to improve spatial consistency. The resulting model, Trans$^2$-CBCT, provides an additional gain of 0.63 dB PSNR and 0.0117 SSIM. Experiments on LUNA16 and ToothFairy show consistent gains from six to ten views, validating the effectiveness of combining CNN-Transformer features with point-based geometry reasoning for sparse-view CBCT reconstruction.

Yue Ma, Huantao Ren, Boyu Wang et al.

Continual learning aims to update a model so that it can sequentially learn new tasks without forgetting previously acquired knowledge. Recent continual learning approaches often leverage the vision-language model CLIP for its high-dimensional feature space and cross-modality feature matching. Traditional CLIP-based classification methods identify the most similar text label for a test image by comparing their embeddings. However, these methods are sensitive to the quality of text phrases and less effective for classes lacking meaningful text labels. In this work, we rethink CLIP-based continual learning and introduce the concept of Label Vector Pool (LVP). LVP replaces text labels with training images as similarity references, eliminating the need for ideal text descriptions. We present three variations of LVP and evaluate their performance on class and domain incremental learning tasks. Leveraging CLIP's high dimensional feature space, LVP learning algorithms are task-order invariant. The new knowledge does not modify the old knowledge, hence, there is minimum forgetting. Different tasks can be learned independently and in parallel with low computational and memory demands. Experimental results show that proposed LVP-based methods outperform the current state-of-the-art baseline by a significant margin of 40.7%.

Weiheng Chai, Brian Testa, Huantao Ren et al.

Deep neural networks are extensively applied to real-world tasks, such as face recognition and medical image classification, where privacy and data protection are critical. Image data, if not protected, can be exploited to infer personal or contextual information. Existing privacy preservation methods, like encryption, generate perturbed images that are unrecognizable to even humans. Adversarial attack approaches prohibit automated inference even for authorized stakeholders, limiting practical incentives for commercial and widespread adaptation. This pioneering study tackles an unexplored practical privacy preservation use case by generating human-perceivable images that maintain accurate inference by an authorized model while evading other unauthorized black-box models of similar or dissimilar objectives, and addresses the previous research gaps. The datasets employed are ImageNet, for image classification, Celeba-HQ dataset, for identity classification, and AffectNet, for emotion classification. Our results show that the generated images can successfully maintain the accuracy of a protected model and degrade the average accuracy of the unauthorized black-box models to 11.97%, 6.63%, and 55.51% on ImageNet, Celeba-HQ, and AffectNet datasets, respectively.

Geethu Joseph, Chen Zhong, M. Cenk Gursoy et al.

We address the problem of sequentially selecting and observing processes from a given set to find the anomalies among them. The decision-maker observes a subset of the processes at any given time instant and obtains a noisy binary indicator of whether or not the corresponding process is anomalous. In this setting, we develop an anomaly detection algorithm that chooses the processes to be observed at a given time instant, decides when to stop taking observations, and declares the decision on anomalous processes. The objective of the detection algorithm is to identify the anomalies with an accuracy exceeding the desired value while minimizing the delay in decision making. We devise a centralized algorithm where the processes are jointly selected by a common agent as well as a decentralized algorithm where the decision of whether to select a process is made independently for each process. Our algorithms rely on a Markov decision process defined using the marginal probability of each process being normal or anomalous, conditioned on the observations. We implement the detection algorithms using the deep actor-critic reinforcement learning framework. Unlike prior work on this topic that has exponential complexity in the number of processes, our algorithms have computational and memory requirements that are both polynomial in the number of processes. We demonstrate the efficacy of these algorithms using numerical experiments by comparing them with state-of-the-art methods.

Jiajing Chen, Burak Kakillioglu, Senem Velipasalar

With the proliferation of Lidar sensors and 3D vision cameras, 3D point cloud analysis has attracted significant attention in recent years. After the success of the pioneer work PointNet, deep learning-based methods have been increasingly applied to various tasks, including 3D point cloud segmentation and 3D object classification. In this paper, we propose a novel 3D point cloud learning network, referred to as Dynamic Point Feature Aggregation Network (DPFA-Net), by selectively performing the neighborhood feature aggregation with dynamic pooling and an attention mechanism. DPFA-Net has two variants for semantic segmentation and classification of 3D point clouds. As the core module of the DPFA-Net, we propose a Feature Aggregation layer, in which features of the dynamic neighborhood of each point are aggregated via a self-attention mechanism. In contrast to other segmentation models, which aggregate features from fixed neighborhoods, our approach can aggregate features from different neighbors in different layers providing a more selective and broader view to the query points, and focusing more on the relevant features in a local neighborhood. In addition, to further improve the performance of the proposed semantic segmentation model, we present two novel approaches, namely Two-Stage BF-Net and BF-Regularization to exploit the background-foreground information. Experimental results show that the proposed DPFA-Net achieves the state-of-the-art overall accuracy score for semantic segmentation on the S3DIS dataset, and provides a consistently satisfactory performance across different tasks of semantic segmentation, part segmentation, and 3D object classification. It is also computationally more efficient compared to other methods.

Geethu Joseph, Chen Zhong, M. Cenk Gursoy et al.

In this paper, we address the anomaly detection problem where the objective is to find the anomalous processes among a given set of processes. To this end, the decision-making agent probes a subset of processes at every time instant and obtains a potentially erroneous estimate of the binary variable which indicates whether or not the corresponding process is anomalous. The agent continues to probe the processes until it obtains a sufficient number of measurements to reliably identify the anomalous processes. In this context, we develop a sequential selection algorithm that decides which processes to be probed at every instant to detect the anomalies with an accuracy exceeding a desired value while minimizing the delay in making the decision and the total number of measurements taken. Our algorithm is based on active inference which is a general framework to make sequential decisions in order to maximize the notion of free energy. We define the free energy using the objectives of the selection policy and implement the active inference framework using a deep neural network approximation. Using numerical experiments, we compare our algorithm with the state-of-the-art method based on deep actor-critic reinforcement learning and demonstrate the superior performance of our algorithm.

Feng Wang, M. Cenk Gursoy, Senem Velipasalar

Deep reinforcement learning (DRL) has recently been used to perform efficient resource allocation in wireless communications. In this paper, the vulnerabilities of such DRL agents to adversarial attacks is studied. In particular, we consider multiple DRL agents that perform both dynamic channel access and power control in wireless interference channels. For these victim DRL agents, we design a jammer, which is also a DRL agent. We propose an adversarial jamming attack scheme that utilizes a listening phase and significantly degrades the users' sum rate. Subsequently, we develop an ensemble policy defense strategy against such a jamming attacker by reloading models (saved during retraining) that have minimum transition correlation.

Fatih Altay, Guillermo Ramon Sanchez, Yanli James et al.

Alzheimer's disease is one of the diseases that mostly affects older people without being a part of aging. The most common symptoms include problems with communicating and abstract thinking, as well as disorientation. It is important to detect Alzheimer's disease in early stages so that cognitive functioning would be improved by medication and training. In this paper, we propose two attention model networks for detecting Alzheimer's disease from MRI images to help early detection efforts at the preclinical stage. We also compare the performance of these two attention network models with a baseline model. Recently available OASIS-3 Longitudinal Neuroimaging, Clinical, and Cognitive Dataset is used to train, evaluate and compare our models. The novelty of this research resides in the fact that we aim to detect Alzheimer's disease when all the parameters, physical assessments, and clinical data state that the patient is healthy and showing no symptoms

Chen Zhong, M. Cenk Gursoy, Senem Velipasalar

Anomaly detection incurs certain sampling and sensing costs and therefore it is of great importance to strike a balance between the detection accuracy and these costs. In this work, we study anomaly detection by considering the detection of threshold crossings in a stochastic time series without the knowledge of its statistics. To reduce the sampling cost in this detection process, we propose the use of hierarchical generative adversarial networks (GANs) to perform nonuniform sampling. In order to improve the detection accuracy and reduce the delay in detection, we introduce a buffer zone in the operation of the proposed GAN-based detector. In the experiments, we analyze the performance of the proposed hierarchical GAN detector considering the metrics of detection delay, miss rates, average cost of error, and sampling ratio. We identify the tradeoffs in the performance as the buffer zone sizes and the number of GAN levels in the hierarchy vary. We also compare the performance with that of a sampling policy that approximately minimizes the sum of average costs of sampling and error given the parameters of the stochastic process. We demonstrate that the proposed GAN-based detector can have significant performance improvements in terms of detection delay and average cost of error with a larger buffer zone but at the cost of increased sampling rates.

Feng Wang, Chen Zhong, M. Cenk Gursoy et al.

As the applications of deep reinforcement learning (DRL) in wireless communications grow, sensitivity of DRL based wireless communication strategies against adversarial attacks has started to draw increasing attention. In order to address such sensitivity and alleviate the resulting security concerns, we in this paper consider a victim user that performs DRL-based dynamic channel access, and an attacker that executes DRLbased jamming attacks to disrupt the victim. Hence, both the victim and attacker are DRL agents and can interact with each other, retrain their models, and adapt to opponents' policies. In this setting, we initially develop an adversarial jamming attack policy that aims at minimizing the accuracy of victim's decision making on dynamic channel access. Subsequently, we devise defense strategies against such an attacker, and propose three defense strategies, namely diversified defense with proportional-integral-derivative (PID) control, diversified defense with an imitation attacker, and defense via orthogonal policies. We design these strategies to maximize the attacked victim's accuracy and evaluate their performances.

Yilan Li, Senem Velipasalar

Recent works have shown that neural networks are vulnerable to carefully crafted adversarial examples (AE). By adding small perturbations to input images, AEs are able to make the victim model predicts incorrect outputs. Several research work in adversarial machine learning started to focus on the detection of AEs in autonomous driving. However, the existing studies either use preliminary assumption on outputs of detections or ignore the tracking system in the perception pipeline. In this paper, we firstly propose a novel distance metric for practical autonomous driving object detection outputs. Then, we bridge the gap between the current AE detection research and the real-world autonomous systems by providing a temporal detection algorithm, which takes the impact of tracking system into consideration. We perform evaluation on Berkeley Deep Drive (BDD) and CityScapes datasets to show how our approach outperforms existing single-frame-mAP based AE detections by increasing 17.76% accuracy of performance.

Chen Zhong, M. Cenk Gursoy, Senem Velipasalar

Anomaly detection is widely applied in a variety of domains, involving for instance, smart home systems, network traffic monitoring, IoT applications and sensor networks. In this paper, we study deep reinforcement learning based active sequential testing for anomaly detection. We assume that there is an unknown number of abnormal processes at a time and the agent can only check with one sensor in each sampling step. To maximize the confidence level of the decision and minimize the stopping time concurrently, we propose a deep actor-critic reinforcement learning framework that can dynamically select the sensor based on the posterior probabilities. We provide simulation results for both the training phase and testing phase, and compare the proposed framework with the Chernoff test in terms of claim delay and loss.

Chen Zhong, Ziyang Lu, M. Cenk Gursoy et al.

To make efficient use of limited spectral resources, we in this work propose a deep actor-critic reinforcement learning based framework for dynamic multichannel access. We consider both a single-user case and a scenario in which multiple users attempt to access channels simultaneously. We employ the proposed framework as a single agent in the single-user case, and extend it to a decentralized multi-agent framework in the multi-user scenario. In both cases, we develop algorithms for the actor-critic deep reinforcement learning and evaluate the proposed learning policies via experiments and numerical results. In the single-user model, in order to evaluate the performance of the proposed channel access policy and the framework's tolerance against uncertainty, we explore different channel switching patterns and different switching probabilities. In the case of multiple users, we analyze the probabilities of each user accessing channels with favorable channel conditions and the probability of collision. We also address a time-varying environment to identify the adaptive ability of the proposed framework. Additionally, we provide comparisons (in terms of both the average reward and time efficiency) between the proposed actor-critic deep reinforcement learning framework, Deep-Q network (DQN) based approach, random access, and the optimal policy when the channel dynamics are known.

Yantao Lu, Senem Velipasalar

There has been significant amount of research work on human activity classification relying either on Inertial Measurement Unit (IMU) data or data from static cameras providing a third-person view. Using only IMU data limits the variety and complexity of the activities that can be detected. For instance, the sitting activity can be detected by IMU data, but it cannot be determined whether the subject has sat on a chair or a sofa, or where the subject is. To perform fine-grained activity classification from egocentric videos, and to distinguish between activities that cannot be differentiated by only IMU data, we present an autonomous and robust method using data from both ego-vision cameras and IMUs. In contrast to convolutional neural network-based approaches, we propose to employ capsule networks to obtain features from egocentric video data. Moreover, Convolutional Long Short Term Memory framework is employed both on egocentric videos and IMU data to capture temporal aspect of actions. We also propose a genetic algorithm-based approach to autonomously and systematically set various network parameters, rather than using manual settings. Experiments have been performed to perform 9- and 26-label activity classification, and the proposed method, using autonomously set network parameters, has provided very promising results, achieving overall accuracies of 86.6\% and 77.2\%, respectively. The proposed approach combining both modalities also provides increased accuracy compared to using only egovision data and only IMU data.

Chen Zhong, M. Cenk Gursoy, Senem Velipasalar

The growing demand on high-quality and low-latency multimedia services has led to much interest in edge caching techniques. Motivated by this, we in this paper consider edge caching at the base stations with unknown content popularity distributions. To solve the dynamic control problem of making caching decisions, we propose a deep actor-critic reinforcement learning based multi-agent framework with the aim to minimize the overall average transmission delay. To evaluate the proposed framework, we compare the learning-based performance with three other caching policies, namely least recently used (LRU), least frequently used (LFU), and first-in-first-out (FIFO) policies. Through simulation results, performance improvements of the proposed framework over these three caching algorithms have been identified and its superior ability to adapt to varying environments is demonstrated.

Yunhan Jia, Yantao Lu, Senem Velipasalar et al.

Neural networks are known to be vulnerable to carefully crafted adversarial examples, and these malicious samples often transfer, i.e., they maintain their effectiveness even against other models. With great efforts delved into the transferability of adversarial examples, surprisingly, less attention has been paid to its impact on real-world deep learning deployment. In this paper, we investigate the transferability of adversarial examples across a wide range of real-world computer vision tasks, including image classification, explicit content detection, optical character recognition (OCR), and object detection. It represents the cybercriminal's situation where an ensemble of different detection mechanisms need to be evaded all at once. We propose practical attack that overcomes existing attacks' limitation of requiring task-specific loss functions by targeting on the `dispersion' of internal feature map. We report evaluation on four different computer vision tasks provided by Google Cloud Vision APIs to show how our approach outperforms existing attacks by degrading performance of multiple CV tasks by a large margin with only modest perturbations.

Chuang Ye, M. Cenk Gursoy, Senem Velipasalar

In this paper, we investigate the problem of power control for streaming variable bit rate (VBR) videos over wireless links. A system model involving a transmitter (e.g., a base station) that sends VBR video data to a receiver (e.g., a mobile user) equipped with a playout buffer is adopted, as used in dynamic adaptive streaming video applications. In this setting, we analyze power control policies considering the following two objectives: 1) the minimization of the transmit power consumption, and 2) the minimization of the transmission completion time of the communication session. In order to play the video without interruptions, the power control policy should also satisfy the requirement that the VBR video data is delivered to the mobile user without causing playout buffer underflow or overflows. A directional water-filling algorithm, which provides a simple and concise interpretation of the necessary optimality conditions, is identified as the optimal offline policy. Following this, two online policies are proposed for power control based on channel side information (CSI) prediction within a short time window. Dynamic programming is employed to implement the optimal offline and the initial online power control policies that minimize the transmit power consumption in the communication session. Subsequently, reinforcement learning (RL) based approach is employed for the second online power control policy. Via simulation results, we show that the optimal offline power control policy that minimizes the overall power consumption leads to substantial energy savings compared to the strategy of minimizing the time duration of video streaming. We also demonstrate that the RL algorithm performs better than the dynamic programming based online grouped water-filling (GWF) strategy unless the channel is highly correlated.

Chuang Ye, M. Cenk Gursoy, Senem Velipasalar

In this paper, wireless video transmission to multiple users under total transmission power and minimum required video quality constraints is studied. In order to provide the desired performance levels to the end-users in real-time video transmissions while using the energy resources efficiently, we assume that power control is employed. Due to the presence of interference, determining the optimal power control is a non-convex problem but can be solved via monotonic optimization framework. However, monotonic optimization is an iterative algorithm and can often entail considerable computational complexity, making it not suitable for real-time applications. To address this, we propose a learning-based approach that treats the input and output of a resource allocation algorithm as an unknown nonlinear mapping and a deep neural network (DNN) is employed to learn this mapping. This learned mapping via DNN can provide the optimal power level quickly for given channel conditions.

Chen Zhong, Ziyang Lu, M. Cenk Gursoy et al.

We consider the dynamic multichannel access problem, which can be formulated as a partially observable Markov decision process (POMDP). We first propose a model-free actor-critic deep reinforcement learning based framework to explore the sensing policy. To evaluate the performance of the proposed sensing policy and the framework's tolerance against uncertainty, we test the framework in scenarios with different channel switching patterns and consider different switching probabilities. Then, we consider a time-varying environment to identify the adaptive ability of the proposed framework. Additionally, we provide comparisons with the Deep-Q network (DQN) based framework proposed in [1], in terms of both average reward and the time efficiency.

Yantao Lu, Burak Kakillioglu, Senem Velipasalar

The choice of parameters, and the design of the network architecture are important factors affecting the performance of deep neural networks. Genetic Algorithms (GA) have been used before to determine parameters of a network. Yet, GAs perform a finite search over a discrete set of pre-defined candidates, and cannot, in general, generate unseen configurations. In this paper, to move from exploration to exploitation, we propose a novel and systematic method that autonomously and simultaneously optimizes multiple parameters of any deep neural network by using a GA aided by a bi-generative adversarial network (Bi-GAN). The proposed Bi-GAN allows the autonomous exploitation and choice of the number of neurons, for fully-connected layers, and number of filters, for convolutional layers, from a large range of values. Our proposed Bi-GAN involves two generators, and two different models compete and improve each other progressively with a GAN-based strategy to optimize the networks during GA evolution. Our proposed approach can be used to autonomously refine the number of convolutional layers and dense layers, number and size of kernels, and the number of neurons for the dense layers; choose the type of the activation function; and decide whether to use dropout and batch normalization or not, to improve the accuracy of different deep neural network architectures. Without loss of generality, the proposed method has been tested with the ModelNet database, and compared with the 3D Shapenets and two GA-only methods. The results show that the presented approach can simultaneously and successfully optimize multiple neural network parameters, and achieve higher accuracy even with shallower networks.

Burak Kakillioglu, Yantao Lu, Senem Velipasalar

The choice of parameters, and the design of the network architecture are important factors affecting the performance of deep neural networks. However, there has not been much work on developing an established and systematic way of building the structure and choosing the parameters of a neural network, and this task heavily depends on trial and error and empirical results. Considering that there are many design and parameter choices, such as the number of neurons in each layer, the type of activation function, the choice of using drop out or not, it is very hard to cover every configuration, and find the optimal structure. In this paper, we propose a novel and systematic method that autonomously and simultaneously optimizes multiple parameters of any given deep neural network by using a generative adversarial network (GAN). In our proposed approach, two different models compete and improve each other progressively with a GAN-based strategy. Our proposed approach can be used to autonomously refine the parameters, and improve the accuracy of different deep neural network architectures. Without loss of generality, the proposed method has been tested with three different neural network architectures, and three very different datasets and applications. The results show that the presented approach can simultaneously and successfully optimize multiple neural network parameters, and achieve increased accuracy in all three scenarios.

Chuang Ye, M. Cenk Gursoy, Senem Velipasalar

In this paper, we investigate the problem of power control for streaming variable-bit-rate (VBR) videos in a device-to-device (D2D) wireless network. A VBR video traffic model that considers video frame sizes and playout buffers at the mobile users is adopted. A setup with one pair of D2D users (DUs) and one cellular user (CU) is considered and three modes, namely cellular mode, dedicated mode and reuse mode, are employed. Mode selection for the data delivery is determined and the transmit powers of the base station (BS) and device transmitter are optimized with the goal of maximizing the overall transmission rate while VBR video data can be delivered to the CU and DU without causing playout buffer underflows or overflows. A low-complexity algorithm is proposed. Through simulations with VBR video traces over fading channels, we demonstrate that video delivery with mode selection and power control achieves a better performance than just using a single mode throughout the transmission.

Mehmet Emin Basbug, Koray Ozcan, Senem Velipasalar

As part of daily monitoring of human activities, wearable sensors and devices are becoming increasingly popular sources of data. With the advent of smartphones equipped with acceloremeter, gyroscope and camera; it is now possible to develop activity classification platforms everyone can use conveniently. In this paper, we propose a fast inference method for an unsupervised non-parametric time series model namely variational inference for sticky HDP-SLDS(Hierarchical Dirichlet Process Switching Linear Dynamical System). We show that the proposed algorithm can differentiate various indoor activities such as sitting, walking, turning, going up/down the stairs and taking the elevator using only the acceloremeter of an Android smartphone Samsung Galaxy S4. We used the front camera of the smartphone to annotate activity types precisely. We compared the proposed method with Hidden Markov Models with Gaussian emission probabilities on a dataset of 10 subjects. We showed that the efficacy of the stickiness property. We further compared the variational inference to the Gibbs sampler on the same model and show that variational inference is faster in one order of magnitude.