Generative self-supervised learning on graphs has emerged as a popular learning paradigm and demonstrated its efficacy in handling non-Euclidean data. However, several remaining issues limit the capability of existing methods: 1) the disregard of uneven node significance in masking, 2) the underutilization of holistic graph information, 3) the ignorance of semantic knowledge in the representation space due to the exclusive use of reconstruction loss in the output space, and 4) the unstable reconstructions caused by the large volume of masked contents. In light of this, we propose ACE-GSL, an adaptive and context-rich graph self-supervised learning framework to address these issues from the perspectives of adaptivity, integrity, complementarity, and consistency. Specifically, we first develop an adaptive feature mask generator to account for the unique significance of nodes and sample informative masks (adaptivity). We then design a ranking-based structure reconstruction objective joint with feature reconstruction to capture holistic graph information and emphasize the topological proximity between neighbors (integrity). After that, we present a bootstrapping-based similarity module to encode the high-level semantic knowledge in the representation space, complementary to the low-level reconstruction in the output space (complementarity). Finally, we build a consistency assurance module to provide reconstruction objectives with extra stabilized consistency targets (consistency). Extensive experiments demonstrate that ACE-GSL achieves state-of-the-art performance over 28 methods on 20 datasets across 3 tasks.

We introduce a real-time system for tracking hand pose using 6- axis inertial measurement units (IMUs) without requiring magnetometers or external sensors. Accurate hand pose tracking with only 6-axis IMUs is known to be fundamentally challenging due to the absence of a shared heading reference, leading to severe drift and inter-sensor misalignment. To overcome these limitations, we propose a hybrid method that combines a learning-based pose estimation approach followed by a late-stage Extended Kalman Filter (EKF). The learning-based model estimates noisy yet reasonable hand poses and is trained with drift-insensitive features like gravity vectors and wrist-relative gyroscope signals. On the other hand the EKF can appropriately filter the noise from pose estimates leading to robust tracking. Evaluated on a 12-hour dataset spanning 23 interaction tasks across 10 participants, our system improves joint angle accuracy by 40% over an EKF-only baseline and by 18% over a learning-only approach, achieving a mean joint error below 10°. The resulting framework enables real-time hand tracking invariant to magnetic perturbations, occlusion, or lighting changes, and is well suited for robotics, human–robot interaction (HRI), and human-computer interaction (HCI) applications.

Current face recognition (FR) algorithms frequently encounter discrimination issues in terms of various attributes (e.g., gender, age) due to the biased demographic distribution of the training datasets towards specific groups. In this paper, we study an identity protected fair FR problem where the goal is to augment the datasets with external face images while ensuring the anonymity of the corresponding face identities. Our problem is motivated by the limitation of current fairness driven data augmentation approaches that directly utilize the external face images accessed by FR algorithm developers while ignoring the protection on the face identities of the image owners. To address the problem, we develop FaDE, a face segment driven identity anonymization framework that augments biased face image datasets by identifying specific face segments with diversified demographic characteristics from external face images but with least identity disclosure, and then reconstructing the segments to full face images with new identities. As a result, the augmented dataset is under a more balanced demographic distribution and improves the fairness performance of the optimized FR algorithms. We evaluate FaDE on two public face datasets, CelebA and LFW that suffer from various demographic imbalance. The results show that FaDE significantly enhances both fairness and accuracy performance of the optimized FR algorithms, while keeping effective anonymity for the identities of external face images.

Text-to-image prompt refinement (T2I-Refine) aims to rephrase or extend an input prompt with more descriptive details that can be leveraged to generate images with higher quality.In this paper, we study an adversarial prompt attacking problem for T2I-Refine, where to goal is to implicitly inject specific concept bias to the input prompts during the refinement process so that the generated images, still with higher quality, are explicitly biased to the target group.Our study is motivated by the limitation of current T2I-Refine research that lacks of explorations on the potential capacity of T2I-Refine models to provide prompt refinement service in a biased or advertising manner.To address the limitations, we develop RAt, a prompt refinement and attacking framework that attacks input prompts with intentionally selected adversarial replacements by optimizing a token distribution matrix based on the text-to-image finetuning strategy with a tokenlevel bias obfuscation loss as regularization.We evaluate RAt on a large-scale text-to-image dataset with various concepts as target in both in-domain and transfer-domain scenarios.The evaluation results demonstrate that, compared to other T2I-Refine schemes, RAt is well capable of implicitly attacking input prompts to generate images with higher quality and explicit visual bias towards specific concept group.

Zero-Shot Story Visualization (ZSV) seeks to depict textual narratives through a sequence of images without relying on pre-existing text-image pairs for training. In this paper, we address the challenge of automated multi-character ZSV, aiming to create distinctive yet compatible character portraits for high-quality story visualization without the need of manual human interventions. Our study is motivated by the limitation of current ZSV approaches that necessitate inefficient manual collection of external images as initial character portraits and suffer from low-quality story visualization, especially with multi-character interactions, when the portraits are not well initiated. To overcome these issues, we develop LeMon, an LLM enhanced Multi-Character Zero-Shot Visualization framework that automates character portrait initialization and supports iterative portrait refinement by exploring the semantic content of the story. In particular, we design an LLM-based portrait generation strategy that matches the story characters with external movie characters, and leverage the matched resources as in-context learning (ICL) samples for LLMs to accurately initialize the character portraits. We then propose a graph-based Text2Image diffusion model that constructs a character interaction graph from the story to iteratively refine the character portraits by maximizing the distinctness of different characters while minimizing their incompatibility in the multi-character story visualization. Our evaluation results show that LeMon outperforms existing ZSV approaches in generating high-quality visualizations for stories across various types with multiple interacted characters. Our code is available at https://github.com/arxrean/LLM-LeMon.

Next Basket Recommendation (NBR) aims to recommend a set of products as a basket to users based on their historical shopping behavior. In this paper, we investigate the problem of NBR in online marketplaces (e.g., Instacart, Uber Eats) that connect users with multiple sellers. In such scenarios, effective NBR can significantly enhance the shopping experience of users by recommending diversified and completed products based on specific sellers, especially when a user purchases from a seller they have not visited before. However, conventional NBR approaches assume that all considered products are from the same sellers, which overlooks the complex relationships between users, sellers, and products. To address such limitations, we develop SecGT, a sequential collaborative graph transformer framework that recommends users with baskets from specific sellers based on seller-aware user preference representations that are generated by collaboratively modeling the joint user-seller-product interactions and sequentially exploring the user-agnostic basket transitions in an interactive way. We evaluate the performance of SecGT on users from a leading online marketplace at multiple cities with various involved sellers. The results show that SecGT outperforms existing NBR and also traditional product recommendation approaches on recommending baskets from cold sellers for different types of users across all cities.

In this paper, we study a controllable prompt adversarial attacking problem for text guided image generation (Text2Image) models in the black-box scenario, where the goal is to attack specific visual subjects (e.g., changing a brown dog to white) in a generated image by slightly, if not imperceptibly, perturbing the characters of the driven prompt (e.g., ``brown'' to ``br0wn''). Our study is motivated by the limitations of current Text2Image attacking approaches that still rely on manual trials to create adversarial prompts. To address such limitations, we develop CharGrad, a character-level gradient based attacking framework that replaces specific characters of a prompt with pixel-level similar ones by interactively learning the perturbation direction for the prompt and updating the attacking examiner for the generated image based on a novel proxy perturbation representation for characters. We evaluate CharGrad using the texts from two public image captioning datasets. Results demonstrate that CharGrad outperforms existing text adversarial attacking approaches on attacking various subjects of generated images by black-box Text2Image models in a more effective and efficient way with less perturbation on the characters of the prompts.

Despite their capacity to convey knowledge, most existing knowledge graphs (KGs) are created for specific domains using low-resource data sources, especially those in non-global languages, and thus unavoidably suffer from the incompleteness problem. The automatic discovery of missing triples for KG completion is thus hindered by the challenging long-tail relations problem in low-resource KGs. Few-shot learning models trained on rich-resource KGs are unable to tackle this challenge due to a lack of generalization. To alleviate the impact of the intractable long-tail problem on low-resource KG completion, in this paper, we propose a novel few-shot learning framework empowered by multi-view task representation generation. The framework consists of four components, i.e., few-shot learner, perturbed few-shot learner, relation knowledge distiller, and pairwise contrastive distiller. The key idea is to utilize the different views of each few-shot task to improve and regulate the training of the few-shot learner. For each few-shot task, instead of augmenting it by complicated task designs, we generate its representation of different views using the relation knowledge distiller and perturbed few-shot learner, which are obtained by distilling knowledge from a KG encoder and perturbing the few-shot learner. Then, the generated representation of different views is utilized by the pairwise contrastive distiller based on a teacher-student framework to distill the knowledge of how to represent relations from different views into the few-shot learner and facilitate few-shot learning. Extensive experiments conducted on several real-world low-resource KGs validate the effectiveness of our proposed method.

Fauxtography is a category of multi-modal posts that spread misleading information on various big data online social platforms that generate billions of posts on a daily basis (e.g., Facebook, Twitter, Reddit). A fauxtography post usually consists of an image, a text description and comments from its readers. In this paper, we focus on explaining fauxtography posts by identifyingwhatspecific component andwhythat component of a post leads to the fauxtography (i.e., duo explanations). This problem is motivated by the limitations of current fauxtography detection solutions that only focus on the detection but ignore the importantexplanationaspect of their results. Two critical challenges exist in solving our problem: i) it is difficult to accurately identify the “guilty” component of a fauxtography post given the fact that different components of the post and their associations could all lead to the fauxtography; ii) it is expensive and time-consuming to obtain a good training set with fine-grained labels of fauxtography posts in terms of explainability, making it challenging to develop fully supervised explainable solutions. To address the above challenges, we develop aDuoExplainableFauxtography Detection Framework under aConstrained Supervision (DExFC) to generate duo explanations from both content and comment parts of fauxtography posts. We evaluate the DExFC by creating real-world datasets from different online social media platforms (Twitter and Reddit). The results show that DExFC not only detects the fauxtography posts more accurately than the state-of-the-art solutions but also provides well-justified explanations to its results without the full supervision.

In the real-world application of COVID-19 misinformation detection, a fundamental challenge is the lack of the labeled COVID data to enable supervised end-to-end training of the models, especially at the early stage of the pandemic. To address this challenge, we propose an unsupervised domain adaptation framework using contrastive learning and adversarial domain mixup to transfer the knowledge from an existing source data domain to the target COVID-19 data domain. In particular, to bridge the gap between the source domain and the target domain, our method reduces a radial basis function (RBF) based discrepancy between these two domains. Moreover, we leverage the power of domain adversarial examples to establish an intermediate domain mixup, where the latent representations of the input text from both domains could be mixed during the training process. Extensive experiments on multiple real-world datasets suggest that our method can effectively adapt misinformation detection systems to the unseen COVID-19 target domain with significant improvements compared to the state-of-the-art baselines.

Despite recent progress in improving the performance of misinformation detection systems, classifying misinformation in an unseen domain remains an elusive challenge. To address this issue, a common approach is to introduce a domain critic and encourage domain-invariant input features. However, early misinformation often demonstrates both conditional and label shifts against existing misinformation data (e.g., class imbalance in COVID-19 datasets), rendering such methods less effective for detecting early misinformation. In this paper, we propose contrastive adaptation network for early misinformation detection (CANMD). Specifically, we leverage pseudo labeling to generate high-confidence target examples for joint training with source data. We additionally design a label correction component to estimate and correct the label shifts (i.e., class priors) between the source and target domains. Moreover, a contrastive adaptation loss is integrated in the objective function to reduce the intra-class discrepancy and enlarge the inter-class discrepancy. As such, the adapted model learns corrected class priors and an invariant conditional distribution across both domains for improved estimation of the target data distribution. To demonstrate the effectiveness of the proposed CANMD, we study the case of COVID-19 early misinformation detection and perform extensive experiments using multiple real-world datasets. The results suggest that CANMD can effectively adapt misinformation detection systems to the unseen COVID-19 target domain with significant improvements compared to the state-of-the-art baselines.

Question answering (QA) has demonstrated impressive progress in answering questions from customized domains. Nevertheless, domain adaptation remains one of the most elusive challenges for QA systems, especially when QA systems are trained in a source domain but deployed in a different target domain. In this work, we investigate the potential benefits of question classification for QA domain adaptation. We propose a novel framework: Question Classification for Question Answering (QC4QA). Specifically, a question classifier is adopted to assign question classes to both the source and target data. Then, we perform joint training in a self-supervised fashion via pseudo-labeling. For optimization, inter-domain discrepancy between the source and target domain is reduced via maximum mean discrepancy (MMD) distance. We additionally minimize intra-class discrepancy among QA samples of the same question class for fine-grained adaptation performance. To the best of our knowledge, this is the first work in QA domain adaptation to leverage question classification with self-supervised adaptation. We demonstrate the effectiveness of the proposed QC4QA with consistent improvements against the state-of-the-art baselines on multiple datasets.

Modern smart sensor-based energy management systems leverage non-intrusive load monitoring (NILM) to predict and optimize appliance load distribution in real-time. NILM, or energy disaggregation, refers to the decomposition of electricity usage conditioned on the aggregated power signals (i.e., smart sensor on the main channel). Based on real-time appliance power prediction using sensory technology, energy disaggregation has great potential to increase electricity efficiency and reduce energy expenditure. With the introduction of transformer models, NILM has achieved significant improvements in predicting device power readings. Nevertheless, transformers are less efficient due to O(l2) complexity w.r.t. sequence length l. Moreover, transformers can fail to capture local signal patterns in sequence-to-point settings due to the lack of inductive bias in local context. In this work, we propose an efficient localness transformer for non-intrusive load monitoring (ELTransformer). Specifically, we leverage normalization functions and switch the order of matrix multiplication to approximate self-attention and reduce computational complexity. Additionally, we introduce localness modeling with sparse local attention heads and relative position encodings to enhance the model capacity in extracting short-term local patterns. To the best of our knowledge, ELTransformer is the first NILM model that addresses computational complexity and localness modeling in NILM. With extensive experiments and quantitative analyses, we demonstrate the efficiency and effectiveness of the the proposed ELTransformer with considerable improvements compared to state-of-the-art baselines.

In this paper, we study an explainable COVID-19 misinformation detection problem where the goal is to accurately identify COVID-19 misleading posts on social media and explain the posts with natural language explanations (NLEs). Our problem is motivated by the limitations of current explainable misinformation detection approaches that cannot provide NLEs for COVID-19 posts due to the lack of sufficient professional COVID-19 knowledge for supervision. To address such a limitation, we develop CEA-COVID, a crowd-expert-AI framework that jointly exploits the common logical reasoning ability of online crowd workers and the professional knowledge of COVID-19 experts to effectively generate NLEs for detecting and explaining COVID-19 misinformation. We evaluate CEA-COVID using two public COVID-19 misinformation datasets on social media. Results demonstrate that CEA-COVID outperforms existing explainable misinformation detection models in terms of both explainability and detection accuracy.

In many applications with real-world consequences, it is crucial to develop reliable uncertainty estimation for the predictions made by the AI decision systems. Targeting at the goal of estimating uncertainty, various deep neural network (DNN) based uncertainty estimation algorithms have been proposed. However, the robustness of the uncertainty returned by these algorithms has not been systematically explored. In this work, to raise the awareness of the research community on robust uncertainty estimation, we show that state-of-the-art uncertainty estimation algorithms could fail catastrophically under our proposed adversarial attack despite their impressive performance on uncertainty estimation. In particular, we aim at attacking out-domain uncertainty estimation: under our attack, the uncertainty model would be fooled to make high-confident predictions for the out-domain data, which they originally would have rejected. Extensive experimental results on various benchmark image datasets show that the uncertainty estimated by state-of-the-art methods could be easily corrupted by our attack.

In this paper, we focus on the quality of information service (QoIS) of COVID-19-related information on social media. Our goal is to provide reliable COVID-19 information service by accurately detecting the misleading COVID-19 posts on social media by exploring the community-contributed COVID-19 fact data (CCFD) from different social media platforms. In particular, CCFD refers to the fact-checking reports that are submitted to each social media platform by its users and fact-checking professionals. Our work is motivated by the observation that CCFD often contains useful COVID-19 knowledge facts (e.g., "COVID-19 is not a flu") that can effectively facilitate the identification of misleading COVID-19 social media posts. However, CCFD is often private to the individual social media platform that owns it due to the data privacy concerns such as data copyright of CCFD and user profile information of CCFD contributors. In this paper, we leverage the CCFD from different social media platforms to accurately detect COVID19 misinformation while effectively protecting the privacy of CCFD. Two critical challenges exist in solving our problem: 1) how to generate privacy-aware COVID-19 knowledge facts from the platform-specific CCFD? 2) How to effectively integrate the privacy-aware COVID-19 knowledge facts from different social media platforms to correctly assess the truthfulness of a COVID19 post? To address these challenges, we develop CoviDKG, a COVID-19 distributed knowledge graph framework that constructs a set of CCFD-based knowledge graphs on individual social media platform and exchanges the privacy-aware COVID19 knowledge facts across different platforms to effectively detect misleading COVID-19 posts. We evaluate CoviDKG on two real-world social media datasets and the results show that CoviDKG achieves significant performance gains compared to state-of-the-art baselines in accurately detecting misleading COVID-19 posts on social media.

While sequential recommender systems achieve significant improvements on capturing user dynamics, we argue that sequential recommenders are vulnerable against substitution-based profile pollution attacks. To demonstrate our hypothesis, we propose a substitution-based adversarial attack algorithm, which modifies the input sequence by selecting certain vulnerable elements and substituting them with adversarial items. In both untargeted and targeted attack scenarios, we observe significant performance deterioration using the proposed profile pollution algorithm. Motivated by such observations, we design an efficient adversarial defense method called Dirichlet neighborhood sampling. Specifically, we sample item embeddings from a convex hull constructed by multi-hop neighbors to replace the original items in input sequences. During sampling, a Dirichlet distribution is used to approximate the probability distribution in the neighborhood such that the recommender learns to combat local perturbations. Additionally, we design an adversarial training method tailored for sequential recommender systems. In particular, we represent selected items with one-hot encodings and perform gradient ascent on the encodings to search for the worst case linear combination of item embeddings in training. As such, the embedding function learns robust item representations and the trained recommender is resistant to test-time adversarial examples. Extensive experiments show the effectiveness of both our attack and defense methods, which consistently outperform baselines by a significant margin across model architectures and datasets.

Human face images represent a rich set of visual information for online social media platforms to optimize the machine learning (ML)/AI models in their data-driven facial applications (e.g., face detection, face recognition). However, there exists a growing privacy concern from social media users to share their online face images that will be annotated by unknown crowd workers and analyzed by ML/AI researchers in the model training and optimization process. In this paper, we focus on a privacy-aware face recognition problem where the goal is to empower the facial applications to train their face recognition models with images shared by social media users while protecting the identity of the users. Our problem is motivated by the limitation of current privacy-aware face recognition approaches that mainly prevent algorithmic attacks by manipulating face images but largely ignore the potential privacy leakage related to human activities (e.g., crowdsourcing annotation). To address such limitations, we develop FaceCrowd, a web crowdsourcing based face partition approach to improve the performance of current face recognition models by designing a novel crowdsourced partial face graph generated from privacy-preserved social media face images. We evaluate the performance of FaceCrowd using two real-world human face datasets that consist of large-scale human face images. The results show that FaceCrowd not only improves the accuracy of the face recognition models but also effectively protects the identity information of the social media users who share their face images.

This paper focuses on a critical problem of explainable multimodal COVID-19 misinformation detection where the goal is to accurately detect misleading information in multimodal COVID-19 news articles and provide the reason or evidence that can explain the detection results. Our work is motivated by the lack of judicious study of the association between different modalities (e.g., text and image) of the COVID-19 news content in current solutions. In this paper, we present a generative approach to detect multimodal COVID-19 misinformation by investigating the cross-modal association between the visual and textual content that is deeply embedded in the multimodal news content. Two critical challenges exist in developing our solution: 1) how to accurately assess the consistency between the visual and textual content of a multimodal COVID-19 news article? 2) How to effectively retrieve useful information from the unreliable user comments to explain the misinformation detection results? To address the above challenges, we develop a duo-generative explainable misinformation detection (DGExplain) framework that explicitly explores the cross-modal association between the news content in different modalities and effectively exploits user comments to detect and explain misinformation in multimodal COVID-19 news articles. We evaluate DGExplain on two real-world multimodal COVID-19 news datasets. Evaluation results demonstrate that DGExplain significantly outperforms state-of-the-art baselines in terms of the accuracy of multimodal COVID-19 misinformation detection and the explainability of detection explanations.

The proliferation of social media has promoted the spread of misinformation that raises many concerns in our society. This paper focuses on a critical problem of explainable COVID-19 misinformation detection that aims to accurately identify and explain misleading COVID-19 claims on social media. Motivated by the lack of COVID-19 relevant knowledge in existing solutions, we construct a novel crowdsource knowledge graph based approach to incorporate the COVID-19 knowledge facts by leveraging the collaborative efforts of expert and non-expert crowd workers. Two important challenges exist in developing our solution: i) how to effectively coordinate the crowd efforts from both expert and non-expert workers to generate the relevant knowledge facts for detecting COVID-19 misinformation; ii) How to leverage the knowledge facts from the constructed knowledge graph to accurately explain the detected COVID-19 misinformation. To address the above challenges, we develop HC-COVID, a hierarchical crowdsource knowledge graph based framework that explicitly models the COVID-19 knowledge facts contributed by crowd workers with different levels of expertise and accurately identifies the related knowledge facts to explain the detection results. We evaluate HC-COVID using two public real-world datasets on social media. Evaluation results demonstrate that HC-COVID significantly outperforms state-of-the-art baselines in terms of the detection accuracy of misleading COVID-19 claims and the quality of the explanations.

Human-centric fauxtography is a category of multi-modal posts that spread misleading information on online information distribution and sharing platforms such as online social media. The reason of a human-centric post being fauxtography is closely related to its multi-modal content that consists of diversified human and non-human subjects with complex and implicit relationships. In this paper, we focus on an explainable fauxtography detection problem where the goal is to accurately identify and explain why a human-centric social media post is fauxtography (or not). Our problem is motivated by the limitations of current fauxtography detection solutions that focus primarily on the detection task but ignore the important aspect of explaining their results (e.g., why a certain component of the post delivers the misinformation). Two important challenges exist in solving our problem: 1) it is difficult to capture the implicit relations and attributions of different subjects in a fauxtography post given the fact that many of such knowledge is shared between different crowd workers; 2) it is not a trivial task to create a multi-modal knowledge graph from crowd workers to identify and explain human-centric fauxtography posts with multi-modal contents. To address the above challenges, we develop CrowdGraph, a crowdsourcing based multi-modal knowledge graph approach to address the explainable fauxtography detection problem. We evaluate the performance of CrowdGraph by creating a real-world dataset that consists of human-centric fauxtography posts from Twitter and Reddit. The results show that CrowdGraph not only detects the fauxtography posts more accurately than the state-of-the-arts but also provides well-justified explanations to the detection results with convincing evidence.

Disaster damage assessment (DDA) has emerged as an important application in disaster response and management, which aims to assess the damage severity of an affected area by leveraging AI (e.g., deep learning) techniques to examine the imagery data posted on social media during a disaster event. In this paper, we focus on a crowd-guided neural architecture searching (NAS) problem in DDA applications. Our goal is to leverage human intelligence from crowdsourcing systems to guide the discovery of the optimal neural network architecture in the design space to achieve the desirable damage assessment performance. Our work is motivated by the limitation that the deep neural network architectures in current DDA solutions are mainly designed by AI experts, which is known to be both time-consuming and error-prone. Two critical technical challenges exist in solving our problem: i) it is challenging to design a manageable NAS space for crowd-based solutions; ii) it is non-trivial to transfer the imperfect crowd knowledge to effective decisions in identifying the optimal neural network architecture of a DDA application. To address the above challenges, we develop CrowdNAS, a crowd-guided NAS framework that develops novel techniques inspired by AI, crowdsourcing, and estimation theory to address the NAS problem. The evaluation results from two real-world DDA applications show that CrowdNAS consistently outperforms the state-of-the-art AI-only, crowd-AI, and NAS baselines by achieving the highest classification accuracy in the damage assessment while maintaining a low computational cost under various evaluation scenarios.

AI-based smart urban sensing (ASUS) has emerged as a scalable and pervasive application paradigm in smart city planning and management that aims to automatically assess the physical status of the urban environments by leveraging AI techniques and massive urban sensing data. In this paper, we focus on a crowd-driven neural network (NN) hyperparameter optimization problem in ASUS applications. Our goal is to utilize the human intelligence from crowdsourcing systems to identify the optimal NN hyperparameter configuration for an ASUS model. Our work is motivated by the observation that the hyperparameters of current ASUS models are often manually configured by the AI specialists, which is known to be both error-prone and suboptimal. Two key technical challenges exist in solving our problem: i) it is challenging to effectively translate the highly complex NN hyperparameter optimization problem in AI to a simplified problem that can be solved by crowd workers without extensive AI expertise; ii) it is difficult to identify the optimal hyperparameter configuration in the large hyperparameter search space given the blackbox nature of the AI model. To address the above challenges, we develop CrowdOptim, a crowd-AI collaborative learning framework that integrates the techniques from crowdsourcing, hyperparameter optimization, and estimation theory to address the crowd-driven NN hyperparameter optimization problem in ASUS applications. The evaluation results from two real-world ASUS applications (i.e., smart city infrastructure monitoring (SCIM) and urban environment cleanliness assessment (UECA)) show that CrowdOptim consistently outperforms the state-of-the-art deep convolutional networks, crowd-AI, and hyperparameter optimization baselines in achieving the ASUS application objectives under various evaluation scenarios.

Cultural heritage sites are precious and fragile resources that hold significant historical, esthetic, and social values in our society. However, the increasing frequency and severity of natural and man-made disasters constantly strike the cultural heritage sites with significant damages. In this article, we focus on a cultural heritage damage assessment (CHDA) problem where the goal is to accurately locate the damaged area of a cultural heritage site using the imagery data posted on social media during a disaster event by exploring the collective strengths of both AI and human intelligence from crowdsourcing systems. Unlike other infrastructure-based solutions, social media platforms provide a more pervasive and scalable solution to acquire timely cultural heritage damage information during disaster events. Our work is motivated by the limitation of current AI solutions that fail to accurately model the complex cultural heritage damage due to the lack of essential human cultural knowledge to differentiate various damage types and identify the actual causes of the damage. Two critical technical challenges exist in solving our problem: 1) it is challenging to effectively detect the problematic cultural heritage damage estimation of AI in the absence of ground truth labels and 2) it is nontrivial to acquire accurate cultural background knowledge from the potentially unreliable crowd workers to effectively address the failure cases of AI. To address the above-mentioned challenges, we develop CollabLearn, an uncertainty-aware crowd-AI collaborative assessment system that explicitly explores the human intelligence from crowdsourcing systems to identify and fix AI failure cases and boost the damage assessment accuracy in CHDA applications. The evaluation results on real-world datasets show that CollabLearn consistently outperforms both the state-of-the-art AI-only and crowd-AI hybrid baselines in accurately assessing the damage of several world-renowned cultural heritage sites in recent disaster events.

Social sensing has emerged as a pervasive and scalable sensing paradigm to obtain timely information of the physical world from "human sensors". In this paper, we study a new extremely-sparse disaster damage assessment (DBA) problem in social sensing. The objective is to automatically assess the damage severity of affected areas in a disaster event by leveraging the imagery data reported on online social media with extremely sparse training data (e.g., only 1% of the data samples have labels). Our problem is motivated by the limitation of current DDA solutions that often require a significant amount of high-quality training data to learn an effective DDA model. We identify two critical challenges in solving our problem: i) it remains to be a fundamental challenge on how to effectively train a reliable DDA model given the lack of sufficient damage severity labels; ii) it is a difficult task to capture the excessive and fine-grained damage-related features in each image for accurate damage assessment. In this paper, we propose ContrastDDA, a deep contrastive learning approach to address the extremely-sparse DDA problem by designing an integrated contrastive and augmentative neural network architecture for accurate disaster damage assessment using the extremely sparse training samples. The evaluation results on two real-world DDA applications demonstrate that ContrastDDA clearly outperforms state-of-the-art deep learning and semi-supervised learning baselines with the highest DDA accuracy under different application scenarios.

Human face images represent a rich set of visual data information that is utilized by various big data driven human facial applications. However, the performance of these applications is usually biased towards the majority demographic group due to the data imbalance issue. In this paper, we focus on a fair human face data exchange problem where the goal is to exchange visual features of human face images between different human face datasets and obtain a set of augmented datasets that improve the fairness and performance of human facial applications. Our problem is motivated by the limitations of current fairness approaches that only focus on a single human face dataset from a particular application and require a large amount of pre-annotated demographic attribute labels to develop fair human facial models. To address these limitations, we develop ExgFair, a crowdsourcing-based fair data exchange framework to generate a set of augmented fair face image datasets by leveraging the crowdsourced demographic attribute labels of human face images. We evaluate ExgFair using a set of real-world human face image datasets with different demographic distributions. The results show that ExgFair not only reduces demographic biases of the datasets but also improves the accuracy of human facial applications trained on the augmented fair datasets.

This paper studies an emerging and important problem of identifying misleading COVID-19 short videos where the misleading content is jointly expressed in the visual, audio, and textual content of videos. Existing solutions for misleading video detection mainly focus on the authenticity of videos or audios against AI algorithms (e.g., deepfake) or video manipulation, and are insufficient to address our problem where most videos are user-generated and intentionally edited. Two critical challenges exist in solving our problem: i) how to effectively extract information from the distractive and manipulated visual content in TikTok videos? ii) How to efficiently aggregate heterogeneous information across different modalities in short videos? To address the above challenges, we develop TikTec, a multimodal misinformation detection framework that explicitly exploits the captions to accurately capture the key information from the distractive video content, and effectively learns the composed misinformation that is jointly conveyed by the visual and audio content. We evaluate TikTec on a real-world COVID- 19 video dataset collected from TikTok. Evaluation results show that TikTec achieves significant performance gains compared to state-of-the-art baselines in accurately detecting misleading COVID-19 short videos.

Social sensing is emerging as an effective and pervasive sensing paradigm to collect timely data and observations from human sensors. This paper focuses on the problem of COVID-19 misinformation detection on social media. Our work is motivated by the lack of COVID-specific knowledge in current misinformation detection solutions, which is critical to assess the truthfulness of social media claims about the emerging COVID-19 disease. In this paper, we leverage human intelligence on a crowdsourcing platform to obtain essential knowledge facts for detecting the COVID-19 misinformation on social media. Two critical challenges exist in solving our problem: i) how to efficiently acquire accurate and timely knowledge that is both inclusive and specific to COVID-19? ii) How to effectively coordinate the efforts from both expert and non-expert workers to detect COVID-19 misinformation? To address these challenges, we develop FakeSens, a social sensing based crowd knowledge graph approach that explicitly explores the knowledge facts specific to COVID-19 and models the reliability of different types of crowd workers to capture the misleading COVID-19 claims. Evaluation results on a real-world dataset show that FakeSens significantly outperforms state-of-the-art baselines in accurately detecting misleading claims of COVID-19 on social media.

Social sensing has emerged as a pervasive and scalable sensing paradigm to collect observations of the physical world from human sensors. A key advantage of social sensing is its infrastructure-free nature. In this paper, we focus on a streaming urban infrastructure monitoring (Streaming UIM) problem in social sensing. The goal is to automatically detect the urban infrastructure damages from the streaming imagery data posted on social media by exploring the collective power of both AI and human intelligence from crowdsourcing systems. Our work is motivated by the limitation of current AI and crowdsourcing solutions that either fail in many critical time-sensitive UIM application scenarios or are not easily generalizable to monitor the damage of different types of urban infrastructures. We identify two critical challenges in solving our problem: i) it is difficult to dynamically integrate AI and crowd intelligence to effectively identify and fix the failure cases of AI solutions; ii) it is non-trivial to obtain accurate human intelligence from unreliable crowd workers in streaming UIM applications. In this paper, we propose StreamCollab, a streaming crowd-AI collaborative system that explores the collaborative intelligence from AI and crowd to solve the streaming UIM problem. The evaluation results on a real-world urban infrastructure imagery dataset collected from social media demonstrate that StreamCollab consistently outperforms both state-of-the-art AI and crowd-AI baselines in UIM accuracy while maintaining the lowest computational cost.

The advance of mobile sensing and edge computing has brought new opportunities for abnormal health detection (AHD) systems where edge devices such as smartphones and wearable sensors are used to collect people's health information and provide early alerts for abnormal health conditions such as stroke and depression. The recent development of federated learning (FL) allows participants to collaboratively train powerful AHD models while keeping their health data private to local devices. This paper targets at addressing a critical challenge of adapting FL to train AHD models, where the participants' health data is highly imbalanced and contains biased class distributions. Existing FL solutions fail to address the class imbalance issue due to the strict privacy requirements of participants as well as the heterogeneous resource constraints of their edge devices. In this work, we propose FedSens, a new FL framework dedicated to address the class imbalance problem in AHD applications with explicit considerations of participant privacy and device resource constraints. We evaluate FedSens using a real-world edge computing testbed on two real-world AHD applications. The results show that FedSens can significantly improve the accuracy of AHD models in the presence of severe class imbalance with low energy cost to the edge devices.

Human face image is a large category of visual information utilized by various human facial data services (e.g., face recognition, face generation, face attribute prediction). However, the quality of data services (QoDS) on human face datasets is usually biased towards the majority demographic group due to the data imbalance issue. In this paper, we focus on a fair human face dataset sampling problem where the goal is to sample a sub-dataset from the original dataset to reduce its bias by leveraging crowd intelligence to infer the demographic labels of face images (e.g., male or female, old or young). Our problem is motivated by the limitations of current fair data sampling solutions that require pre-annotated demographic labels to sample a fair dataset. Two important challenges exist in solving our problem: 1) it is extremely time-consuming and expensive to assign crowd workers to annotate demographic labels of all images in a large-scale facial dataset; 2) it is not a trivial task to improve the fairness of the sampled sub-dataset (with fewer data samples) without sacrificing the accuracy performance of data services on such dataset. To address the above challenges, we develop FairCrowd, a fair crowdsourcing-based data sampling framework that leverages an efficient batch-level demographic label inference model and a joint fair-accuracy-aware data shuffling method. We evaluate the performance of FairCrowd through a large-scale real-world face image dataset that consists of celebrity faces from a diversified set of demographic groups. The results show that FairCrowd not only reduces demographic bias but also improves the accuracy of data services trained on the sub-dataset generated by FairCrowd, leading to a more desirable QoDS of the application.

Streaming disaster damage assessment (DDA) aims to automatically assess the damage severity of affected areas in a disaster event on the fly by leveraging the streaming imagery data about the disaster on social media. In this paper, we focus on a dynamic optimal neural architecture searching (NAS) problem in streaming DDA applications. Our goal is to dynamically determine the optimal neural network architecture that accurately estimates the damage severity for each newly-arrived image in the stream by leveraging human intelligence from the crowdsourcing systems. Our work is motivated by the observations that the neural network architectures in current DDA solutions are mainly designed by AI experts, which often leads to non-negligible costs and errors given the dynamic nature of the streaming DDA applications and the lack of real-time annotations of the massive social media data inputs. In this paper, we develop CD-NAS, a crowd-driven dynamic NAS framework that is inspired by novel techniques from AI, crowdsourcing, and estimation theory to address the dynamic optimal NAS problem. The evaluation results from a real-world streaming DDA application show that CD-NAS consistently outperforms the state-of-the-art AI and NAS baselines by achieving the highest disaster damage assessment accuracy.

Weakly Supervised Object Localization (WSOL) methods generate both classification and localization results by learning from only image category labels. Previous methods usually utilize class activation map (CAM) to obtain target object regions. However, most of them only focus on improving foreground object parts in CAM, but ignore the important effect of its background contents. In this paper, we propose a confidence segmentation (ConfSeg) module that builds confidence score for each pixel in CAM without introducing additional hyper-parameters. The generated sample-specific confidence mask is able to indicate the extent of determination for each pixel in CAM, and further supervises additional CAM extended from internal feature maps. Besides, we introduce Co-supervised Augmentation (CoAug) module to capture feature-level representation for foreground and background parts in CAM separately. Then a metric loss is applied at batch sample level to augment distinguish ability of our model, which helps a lot to localize more related object parts. Our final model, CSoA, combines the two modules and achieves superior performance, e.g. 37.69% and 48.81% Top-1 localization error on CUB-200 and ILSVRC datasets, respectively, which outperforms all previous methods and becomes the new state-of-the-art.

Understanding web instructional videos is an essential branch of video understanding in two aspects. First, most existing video methods focus on short-term actions for a-few-second-long video clips; these methods are not directly applicable to long videos. Second, unlike unconstrained long videos, e.g., movies, instructional videos are more structured in that they have step-by-step procedures constraining the understanding task. In this work, we study problem-solving on instructional videos via Visual Question Answering (VQA). Surprisingly, it has not been an emphasis for the video community despite its rich applications. We thereby introduce YouCookQA, an annotated QA dataset for instructional videos based on YouCook2 [27]. The questions in YouCookQA are not limited to cues on a single frame but relations among multiple frames in the temporal dimension. Observing the lack of effective representations for modeling long videos, we propose a set of carefully designed models including a Recurrent Graph Convolutional Network (RGCN) that captures both temporal order and relational information. Furthermore, we study multiple modalities including descriptions and transcripts for the purpose of boosting video understanding. Extensive experiments on YouCookQA suggest that RGCN performs the best in terms of QA accuracy and better performance is gained by introducing human-annotated descriptions. YouCookQA dataset is available at https://github.com/Jossome/YoucookQA.

Fauxtography is a category of multi-modal posts that spreads misleading information on various online social platforms (e.g., Facebook, Twitter, Reddit). A fauxtography post usually consists of an image, a text description and comments from its readers. In this paper, we focus on an explainable fauxtography detection problem where the goal is to explain which a specific component of a post leads to the fauxtography decision. This problem is motivated by the limitations of current fauxtography detection solutions that only focus on the detection but ignore the important explanation aspect of their results. Two critical challenges exist in solving our problem: i) it is difficult to accurately identify the "guilty" component of a fauxtography post given the fact that different components of the post and their associations could all lead to the fauxtography; ii) it is expensive and time-consuming to obtain a good training set with fine-grained labels of fauxtography posts in terms of explainability, making the corresponding solutions weakly supervised in nature. To address the above challenges, we develop ExFaux, an end-to-end graph-based fauxtography explanation framework, to effectively explain which part of the post contributes to its fauxtography. We evaluate the ExFaux by creating a real-world dataset from online social media (Twitter and Reddit). The results show that ExFaux not only detects the fauxtography posts more accurately than the state-of-the-arts but also provides well-justified explanations to its results.

The recent advance of the federated learning (FL) has brought new opportunities for privacy-aware distributed machine learning (ML) applications to train a powerful ML model without accessing the private training data of the participants. In this paper, we focus on addressing a novel fair classification problem in FL where the model trained by FL displays discriminatory bias towards particular demographic groups. Addressing the fairness issue in a FL framework posts three critical challenges: fairness and performance trade-offs, restricted information, and constrained coordination. To address these challenges, we develop FairFL, a fair federated learning framework dedicated to reducing the bias in privacy-sensitive ML applications. It consists of a principled deep multi-agent reinforcement learning framework and a secure information aggregation protocol that optimizes both the accuracy and the fairness of the learned model while respecting the strict privacy constraints of the clients. Evaluation results on real-world applications showed that FairFL can achieve significant performance gains in both fairness and accuracy of the learned model compared to state-of-the-art baselines.