We present Answer-Me, a task-aware multi-task framework which unifies a variety of question answering tasks, such as, visual question answering, visual entailment, visual reasoning. In contrast to previous works using contrastive or generative captioning training, we propose a novel and simple recipe to pre-train a vision-language joint model, which is multi-task as well. The pre-training uses only noisy image captioning data, and is formulated to use the entire architecture end-to-end with both a strong language encoder and decoder. Our results show state-of-the-art performance, zero-shot generalization, robustness to forgetting, and competitive single-task results across a variety of question answering tasks. Our multi-task mixture training learns from tasks of various question intents and thus generalizes better, including on zero-shot vision-language tasks. We conduct experiments in the challenging multi-task and open-vocabulary settings and across a variety of datasets and tasks, such as VQA2.0, SNLI-VE, NLVR2, GQA. We observe that the proposed approach is able to generalize to unseen tasks and that more diverse mixtures lead to higher accuracy in both known and novel tasks.

我们提出了Findit，这是一个简单而多功能的框架，统一了各种视觉接地和本地化任务，包括引用表达理解，基于文本的本地化和对象检测。我们体系结构的关键是一个有效的多尺度融合模块，该模块统一了整个任务中不同的本地化要求。此外，我们发现标准对象检测器在统一这些任务的无需特定任务设计，损失或预计算检测方面非常有效。我们的端到端可训练框架灵活，准确地响应了零，一个或多个对象的广泛的参考表达，本地化或检测查询。在这些任务上进行了共同培训，发现在引用表达和基于文本的本地化方面，胜过最高的艺术状态，并在对象检测中表现出竞争性的性能。最后，与强大的单任务基准相比，Findit可以更好地推广到分布数据和新型类别。所有这些都是通过一个单一的，统一和有效的模型来完成的。代码将发布。

Self-attention has recently been adopted for a wide range of sequence modeling problems. Despite its effectiveness, self-attention suffers from quadratic compute and memory requirements with respect to sequence length. Successful approaches to reduce this complexity focused on attending to local sliding windows or a small set of locations independent of content. Our work proposes to learn dynamic sparse attention patterns that avoid allocating computation and memory to attend to content unrelated to the query of interest. This work builds upon two lines of research: it combines the modeling flexibility of prior work on content-based sparse attention with the efficiency gains from approaches based on local, temporal sparse attention. Our model, the Routing Transformer, endows selfattention with a sparse routing module based on online k-means while reducing the overall complexity of attention to O(n 1.5 d) from O(n 2 d) for sequence length n and hidden dimension d. We show that our model outperforms comparable sparse attention models on language modeling on Wikitext-103 (15.8 vs 18.3 perplexity), as well as on image generation on ImageNet-64 (3.43 vs 3.44 bits/dim) while using fewer self-attention layers. Additionally, we set a new state-of-the-art on the newly released PG-19 data-set, obtaining a test perplexity of 33.2 with a 22 layer Routing Transformer model trained on sequences of length 8192. We open-source the code for Routing Transformer in Tensorflow. *

Deep neural networks (DNNs) are vulnerable to a class of attacks called "backdoor attacks", which create an association between a backdoor trigger and a target label the attacker is interested in exploiting. A backdoored DNN performs well on clean test images, yet persistently predicts an attacker-defined label for any sample in the presence of the backdoor trigger. Although backdoor attacks have been extensively studied in the image domain, there are very few works that explore such attacks in the video domain, and they tend to conclude that image backdoor attacks are less effective in the video domain. In this work, we revisit the traditional backdoor threat model and incorporate additional video-related aspects to that model. We show that poisoned-label image backdoor attacks could be extended temporally in two ways, statically and dynamically, leading to highly effective attacks in the video domain. In addition, we explore natural video backdoors to highlight the seriousness of this vulnerability in the video domain. And, for the first time, we study multi-modal (audiovisual) backdoor attacks against video action recognition models, where we show that attacking a single modality is enough for achieving a high attack success rate.

Unmanned aerial vehicle (UAV) swarms are considered as a promising technique for next-generation communication networks due to their flexibility, mobility, low cost, and the ability to collaboratively and autonomously provide services. Distributed learning (DL) enables UAV swarms to intelligently provide communication services, multi-directional remote surveillance, and target tracking. In this survey, we first introduce several popular DL algorithms such as federated learning (FL), multi-agent Reinforcement Learning (MARL), distributed inference, and split learning, and present a comprehensive overview of their applications for UAV swarms, such as trajectory design, power control, wireless resource allocation, user assignment, perception, and satellite communications. Then, we present several state-of-the-art applications of UAV swarms in wireless communication systems, such us reconfigurable intelligent surface (RIS), virtual reality (VR), semantic communications, and discuss the problems and challenges that DL-enabled UAV swarms can solve in these applications. Finally, we describe open problems of using DL in UAV swarms and future research directions of DL enabled UAV swarms. In summary, this survey provides a comprehensive survey of various DL applications for UAV swarms in extensive scenarios.

Compared to regular cameras, Dynamic Vision Sensors or Event Cameras can output compact visual data based on a change in the intensity in each pixel location asynchronously. In this paper, we study the application of current image-based SLAM techniques to these novel sensors. To this end, the information in adaptively selected event windows is processed to form motion-compensated images. These images are then used to reconstruct the scene and estimate the 6-DOF pose of the camera. We also propose an inertial version of the event-only pipeline to assess its capabilities. We compare the results of different configurations of the proposed algorithm against the ground truth for sequences of two publicly available event datasets. We also compare the results of the proposed event-inertial pipeline with the state-of-the-art and show it can produce comparable or more accurate results provided the map estimate is reliable.

With Twitter's growth and popularity, a huge number of views are shared by users on various topics, making this platform a valuable information source on various political, social, and economic issues. This paper investigates English tweets on the Russia-Ukraine war to analyze trends reflecting users' opinions and sentiments regarding the conflict. The tweets' positive and negative sentiments are analyzed using a BERT-based model, and the time series associated with the frequency of positive and negative tweets for various countries is calculated. Then, we propose a method based on the neighborhood average for modeling and clustering the time series of countries. The clustering results provide valuable insight into public opinion regarding this conflict. Among other things, we can mention the similar thoughts of users from the United States, Canada, the United Kingdom, and most Western European countries versus the shared views of Eastern European, Scandinavian, Asian, and South American nations toward the conflict.

The performance of the Deep Learning (DL) models depends on the quality of labels. In some areas, the involvement of human annotators may lead to noise in the data. When these corrupted labels are blindly regarded as the ground truth (GT), DL models suffer from performance deficiency. This paper presents a method that aims to learn a confident model in the presence of noisy labels. This is done in conjunction with estimating the uncertainty of multiple annotators. We robustly estimate the predictions given only the noisy labels by adding entropy or information-based regularizer to the classifier network. We conduct our experiments on a noisy version of MNIST, CIFAR-10, and FMNIST datasets. Our empirical results demonstrate the robustness of our method as it outperforms or performs comparably to other state-of-the-art (SOTA) methods. In addition, we evaluated the proposed method on the curated dataset, where the noise type and level of various annotators depend on the input image style. We show that our approach performs well and is adept at learning annotators' confusion. Moreover, we demonstrate how our model is more confident in predicting GT than other baselines. Finally, we assess our approach for segmentation problem and showcase its effectiveness with experiments.

This paper deals with the problem of statistical and system heterogeneity in a cross-silo Federated Learning (FL) framework where there exist a limited number of Consumer Internet of Things (CIoT) devices in a smart building. We propose a novel Graph Signal Processing (GSP)-inspired aggregation rule based on graph filtering dubbed ``G-Fedfilt''. The proposed aggregator enables a structured flow of information based on the graph's topology. This behavior allows capturing the interconnection of CIoT devices and training domain-specific models. The embedded graph filter is equipped with a tunable parameter which enables a continuous trade-off between domain-agnostic and domain-specific FL. In the case of domain-agnostic, it forces G-Fedfilt to act similar to the conventional Federated Averaging (FedAvg) aggregation rule. The proposed G-Fedfilt also enables an intrinsic smooth clustering based on the graph connectivity without explicitly specified which further boosts the personalization of the models in the framework. In addition, the proposed scheme enjoys a communication-efficient time-scheduling to alleviate the system heterogeneity. This is accomplished by adaptively adjusting the amount of training data samples and sparsity of the models' gradients to reduce communication desynchronization and latency. Simulation results show that the proposed G-Fedfilt achieves up to $3.99\% $ better classification accuracy than the conventional FedAvg when concerning model personalization on the statistically heterogeneous local datasets, while it is capable of yielding up to $2.41\%$ higher accuracy than FedAvg in the case of testing the generalization of the models.

Learning models are highly dependent on data to work effectively, and they give a better performance upon training on big datasets. Massive research exists in the literature to address the dataset adequacy issue. One promising approach for solving dataset adequacy issues is the data augmentation (DA) approach. In DA, the amount of training data instances is increased by making different transformations on the available data instances to generate new correct and representative data instances. DA increases the dataset size and its variability, which enhances the model performance and its prediction accuracy. DA also solves the class imbalance problem in the classification learning techniques. Few studies have recently considered DA in the Arabic language. These studies rely on traditional augmentation approaches, such as paraphrasing by using rules or noising-based techniques. In this paper, we propose a new Arabic DA method that employs the recent powerful modeling technique, namely the AraGPT-2, for the augmentation process. The generated sentences are evaluated in terms of context, semantics, diversity, and novelty using the Euclidean, cosine, Jaccard, and BLEU distances. Finally, the AraBERT transformer is used on sentiment classification tasks to evaluate the classification performance of the augmented Arabic dataset. The experiments were conducted on four sentiment Arabic datasets, namely AraSarcasm, ASTD, ATT, and MOVIE. The selected datasets vary in size, label number, and unbalanced classes. The results show that the proposed methodology enhanced the Arabic sentiment text classification on all datasets with an increase in F1 score by 4% in AraSarcasm, 6% in ASTD, 9% in ATT, and 13% in MOVIE.