Deep learning models for critical vision tasks remain vulnerable to adversarial attacks.We present SPARC, the first superpixeltargeted black-box attack offering high interpretability and strong performance.Our method uses regional confidence maps to guide perturbations to the most important regions and controls their magnitude using L2 and L1 constraints, keeping changes small and spatially coherent.In targeted attacks, SPARC achieves a competitive success rate with subtle perturbations, while in untargeted attacks, it attains the highest success among superpixel-based and score-based methods with the fewest blackbox queries.SPARC provides a practical balance of performance and interpretability, suitable for real-world black-box scenarios.

Code summarization generates natural language descriptions for code snippets, enhancing readability and maintainability.While current methods perform well with large-scale datasets, they struggle in low-resource scenarios typical of smaller and newer projects.Additionally, developers need summaries that capture project-specific characteristics rather than generic descriptions.To address these challenges, we propose a meta-learning-based training framework that adapts the model to individual projects as distinct tasks, even with minimal data.We introduce a strategy for selecting support projects to boost the framework's effectiveness.Experiments on eight real-world projects show that our method outperforms the baseline approach.Furthermore, we use explainability techniques to clarify the prediction process and identify potential issues.

Visual counterfactual explanations are ideal hypothetical images that change the decision-making of the classifier with high confidence toward the desired class while remaining visually plausible and close to the initial image. In this paper, we propose a new approach to tackle two key challenges in recent prominent works: i) determining which specific counterfactual features are crucial for distinguishing the “concept” of the target class from the original class, and ii) supplying valuable explanations for the non-robust classifier without relying on the support of an adversarially robust model. Our method identifies the essential region for modification through algorithms that provide visual explanations, and then our framework generates realistic counterfactual explanations by combining adversarial attacks based on pruning the adversarial gradient of the target classifier and the latent diffusion model. The proposed method outperforms previous state-of-the-art results on various evaluation criteria on ImageNet and CelebA-HQ datasets. In general, our method can be applied to arbitrary classifiers, highlight the strong association between visual and counterfactual explanations, make semantically meaningful changes from the target classifier, and provide observers with subtle counterfactual images.

Temporal Knowledge Graphs (TKGs) organize dynamic real-world facts, adding a time dimension to the multi-relational graph structure of Knowledge Graphs (KGs).We leverage the expressive power of graph convolutional networks (GCNs) for modeling TKGs, recognizing similarities with handling graph-structured data and utilizing complex geometry.Our approach emphasizes compositional interactions between relations and entities, integrating a diachronic mechanism to enhance representation with both graph structure and temporal dynamics.Experimental results on benchmark datasets, employing various composition operators, showcase the effectiveness of our model in link prediction tasks.

Affordance detection and pose estimation are of great importance in many robotic applications. Their combination helps the robot gain an enhanced manipulation capability, in which the generated pose can facilitate the corresponding affordance task. Previous methods for affodance-pose joint learning are limited to a predefined set of affordances, thus limiting the adaptability of robots in real-world environments. In this paper, we propose a new method for language-conditioned affordance-pose joint learning in 3D point clouds. Given a 3D point cloud object, our method detects the affordance region and generates appropriate 6-DoF poses for any unconstrained affordance label. Our method consists of an open-vocabulary affordance detection branch and a language-guided diffusion model that generates 6-DoF poses based on the affordance text. We also introduce a new high-quality dataset for the task of language-driven affordance-pose joint learning. Intensive experimental results demonstrate that our proposed method works effectively on a wide range of open-vocabulary affordances and outperforms other baselines by a large margin. In addition, we illustrate the usefulness of our method in real-world robotic applications. Our code and dataset are publicly available at https://3DAPNet.github.io.

In Class-Incremental Learning (Class-IL), deep neural networks often fail to learn a sequence of classes incrementally due to catastrophic forgetting, a phenomenon arising from the absence of exposure to old knowledge. To alleviate this issue, conventional rehearsal methods , such as experience replay, store a limited number of old exemplars and then interleave with the current data for joint learning and rehearsal. However, the networks following this training scheme might not successfully reduce forgetting due to the lack of direct consideration of relations between samples of previously learned and new classes. Drawing inspiration from how humans learn by noticing the similarities and differences between classes, we propose a novel Class-IL framework called Relational Replay (RR). RR learns and recalls relations between images across all classes over time. To ensure these relations remain intrinsic and robust to forgetting, we incorporate causal reasoning to RR, resulting in Causal Relational Replay (CRR). CRR analyzes these relations using a causality perspective, aiming to identify intrinsic relations rooted in the images’ semantic features, serving as the cause of these relations. Our proposed method shows a competitive performance compared to the state-of-the-art rehearsal methods in Class-IL with clear and consistent improvements in the majority of settings on standard benchmark datasets.

Due to the global trend towards urbanization, people increasingly move to and live in cities that then continue to grow. Traffic forecasting plays an important role in the intelligent transportation systems of cities as well as in spatio-temporal data mining. State-of-the-art forecasting is achieved by deep-learning approaches due to their ability to contend with complex spatio-temporal dynamics. However, existing methods assume the input is fixed-topology road networks and static traffic time series. These assumptions fail to align with urbanization, where time series are collected continuously and road networks evolve over time. In such settings, deep-learning models require frequent re-initialization and re-training, imposing high computational costs. To enable much more efficient training without jeopardizing model accuracy, we propose the Topological Evolution-aware Framework (TEAM) for traffic forecasting that incorporates convolution and attention. This combination of mechanisms enables better adaptation to newly collected time series while being able to maintain learned knowledge from old time series. TEAM features a continual learning module based on the Wasserstein metric that acts as a buffer that can identify the most stable and the most changing network nodes. Then, only data related to stable nodes is employed for re-training when consolidating a model. Further, only data of new nodes and their adjacent nodes as well as data pertaining to changing nodes are used to re-train the model. Empirical studies with two real-world traffic datasets offer evidence that TEAM is capable of much lower re-training costs than existing methods are, without jeopardizing forecasting accuracy.

Knowledge graph link prediction has been researched for many years.With the steady development of data, the demand for missing link prediction in knowledge bases is growing.In this study, we propose FouriER, a model using Fourier transforms integrated into MetaFormer architecture to learn features from embeddings better but more computationally cost-effective than the self-attention mechanism in Transformer models.Furthermore, we transform embeddings to a 2D form and stack them that benefit the model in learning interactions between entities and relations more efficiently.As a result, we found that our model outperformed baseline models on two benchmark datasets in our experiments.

Despite significant progress over the past few years, ambiguity is still a key challenge in Facial Expression Recognition (FER). It can lead to noisy and inconsistent annotation, which hinders the performance of deep learning models in real-world scenarios. In this paper, we propose a new uncertainty-aware label distribution learning method to improve the robustness of deep models against uncertainty and ambiguity. We leverage neighborhood information in the valence-arousal space to adaptively construct emotiona distributions for training samples. We also consider the uncertainty of provided labels when incorporating them into the label distributions. Our method can be easily integrated into a deep network to obtain more training supervision and improve recognition accuracy. Intensive experiments on several datasets under various noisy and ambiguous settings show that our method achieves competitive results and outperforms recent state-of-the-art approaches. Our code and models are available at https://github.com/minhnhatvt/label-distribution-learning-fer-tf.

Abstract Human emotion recognition is an active research area in artificial intelligence and has made substantial progress over the past few years. Many recent works mainly focus on facial regions to infer human affection, while the surrounding context information is not effectively utilized. In this paper, we proposed a new deep network to effectively recognize human emotions using a novel global-local attention mechanism. Our network is designed to extract features from both facial and context regions independently, then learn them together using the attention module. In this way, both the facial and contextual information is used to infer human emotions, therefore enhancing the discrimination of the classifier. The intensive experiments show that our method surpasses the current state-of-the-art methods on recent emotion datasets by a fair margin. Qualitatively, our global-local attention module can extract more meaningful attention maps than previous methods. The source code and trained model of our network are available at https://github.com/minhnhatvt/glamor-net .

Identifying features to represent graphs such as social networks, protein graphs is increasingly common in both research and business communities, thanks to the fact that data has increased not only in quantity but also in complexity. This results in the graphs to be sparser because not all nodes a re fully connected. In addition, if this whole graph is used as input data for learning algorithms e.g. neural network, a lot of training time will be required. Substantial efforts have been made to convert the graphs to better yet compact representations, among of which is graph embedding. The traditional methods used to map the original graph to its embedding representation had not yielded significant results until deep learning was invented. Many good approaches in this direction, as examples, are DeepWalk, node2vec. However, their general weakness is many important connections in the original graph could be lost. In this paper, we propose another approach to retain more edge information while ensuring the embedding graph is still sufficiently small, compared to the original one. Our experiment results show that the method also increases the accuracy of latter learning models.

Mining High Average-Utility Itemsets (HAUIs) in a quantitative database is an extension of the traditional problem of frequent itemset mining, having several practical applications. Discovering HAUIs is more challenging than mining frequent itemsets using the traditional support model since the average-utilities of itemsets do not satisfy the downward-closure property. To design algorithms for mining HAUIs that reduce the search space of itemsets, prior studies have proposed various upper-bounds on the average-utilities of itemsets. However, these algorithms can generate a huge amount of unpromising HAUI candidates, which result in high memory consumption and long runtimes. To address this problem, this paper proposes four tight average-utility upper-bounds, based on a vertical database representation, and three efficient pruning strategies. Furthermore, a novel generic framework for comparing average-utility upper-bounds is presented. Based on these theoretical results, an efficient algorithm named dHAUIM is introduced for mining the complete set of HAUIs. dHAUIM represents the search space and quickly compute upper-bounds using a novel IDUL structure. Extensive experiments show that dHAUIM outperforms four state-of-the-art algorithms for mining HAUIs in terms of runtime on both real-life and synthetic databases. Moreover, results show that the proposed pruning strategies dramatically reduce the number of candidate HAUIs.

A solution to deploy object detection systems to practical applications is to build cascade frameworks which do threshold comparisons in each stage to efficiently discard a large number of negative objects. For particular applications, these thresholds should be retrained for better effectiveness and the efficiency via training datasets. It means that we have to store labeled datasets permanently or collect huge data (for high-quality thresholds) whenever learning new thresholds. Both approaches are inconvenient and expensive in terms of memory and data collection cost. In this paper, we propose a novel threshold selection mechanism, named Adaptive Cascade Threshold Learning (ACTL), which learns thresholds directly from non-object regions in a single input image instead of object regions from large training data as other existing methods. As a result, we can completely remove the need of training data for cascade threshold learning. Experimental results on two problems of object detection and face detection confirm that our proposed method can obtain the same level of accuracy and speed as state-of-the-art cascade DPM methods but it has the benefit of no threshold training data.

Cascade Deformable Part Models (DPMs) are cascade frameworks to speed up Deformable Part Models (DPMs), which are one of the state-of-the-art solutions for object detection. Its idea is to reject most non-object hypotheses from the early stages of detection process. By investigating the dependency between hypotheses over scales, we introduce a novel pruning method to accelerate Cascade DPM frameworks. Our scale pruning method includes two following strategies: a) rejecting the neighboring scales if the score at a certain scale is too low; b) keeping a few scales with the highest scores, instead of storing all scales per location. Next, we extend this pruning technique to the 3D neighborhood pruning and describe a novel approach to evaluate the root score efficiently without full computation as existing cascade DPM methods. Finally, look-up tables are introduced to work with flexible neighborhood whose volume varies over hypotheses. As a result, our cascade model is equipped with an efficient and aggressive pruning mechanism. Extensive experiments reveal that the proposed method is faster than the state-of-the-art methods for both problems of object detection and face detection.

Frequent pattern mining generates a lot of candidates which spends a lot of usage memory and mining time. Besides, in real applications, a small number of frequent patterns are used. Therefore, the problem of mining top-rank-k frequent patterns (TRFPs) is an interesting topic in recent years. This paper proposes iNTK algorithm for mining TRFPs. This algorithm employs N-list structure generated by PPC-tree to reduce the memory usage. Besides, the subsume concept is also used to enhance the process of mining TRFPs. The experimental results show that iNTK outperforms NTK in terms of mining time and memory usage.

Emergence of novel techniques devices e.g., MS Kinect, enables reliable extraction of human skeletons from action videos. Taking skeleton data as inputs, we propose an approach to extract the discriminative patterns for efficient human action recognition. Each action is considered to consist of a sequence of unit actions, each of which is represented by a pattern. Given a skeleton sequence, we first automatically extract the key-frames, and then categorize them into different patterns. We further use a statistical metric to evaluate the discriminative capability of patterns, and define them as local features for action recognition. Experimental results show that the extracted local descriptors could provide very high accuracy in the action recognition, which demonstrate the efficiency of our method in extracting discriminative unit actions.

This paper presents an effective feature descriptor that integrates intensity order and textures in multi support regions into a compact vector. We first propose the novel Intensity Order Local Binary Pattern (IO-LBP) to encode the texture around each point in an interest region, divide the region according to pixel intensity orders, and then pool patterns to these segments. The IO-LBP descriptor is built by concatenating histograms of all segments together. Besides, multi support regions are used to further improve the discriminative ability. The proposed descriptor can effectively capture both local and global information of an interest region and thus high performance is expected. We evaluate IO-LBP on the standard Oxford dataset and additional images of shadows. Experimental results show that our method is not only invariant to common photometric and geometric transformations, such as illumination change, image rotation, but also robust to complex illumination effects caused by shadows. A significant improvement in performance, comparing to state-of-the-art descriptors, is achieved by IO-LBP.

This paper proposes an algorithm to hide quickly high utility sensitive itemsets. To reduce the execution time, when an item from the sensitive itemset is selected, the rated value is calculated from each transaction containing this item and all values which need be hidden are modified at the same time. Experimental results show that the proposed algorithm improves significantly the runtime.

With the development of the technology like 3D specialized markers, we could capture the moving signals from marker joints and create a huge set of 3D action MoCap data. The more we understand the human action, the better we could apply it to applications like security, analysis of sports, game etc. In order to find the semantically representative features of human actions, we extract the sets of action characteristics which appear frequently in the database. We then propose an Apriori-like algorithm to automatically extract the common sets shared by different action classes. The extracted representative action characteristics are defined in the semantic level, so that it better describes the intrinsic differences between various actions. In our experiments, we show that the knowledge extracted by this method achieves high accuracy of over 80% in recognizing actions on both training and testing data.

There have recently been many methods proposed for matching face sequences in the field of face retrieval. However, most of them have proven to be inefficient in large-scale video databases because they frequently require a huge amount of computational cost to obtain a high degree of accuracy. We present an efficient matching method that is based on the face sequences (called face tracks) in large-scale video databases. The key idea is how to capture the distribution of a face track in the fewest number of low-computational steps. In order to do that, each face track is represented by a vector that approximates the first principal component of the face track distribution and the similarity of face tracks bases on the similarity of these vectors. Our experimental results from a large-scale database of 457,320 human faces extracted from 370 hours of TRECVID videos from 2004-2006 show that the proposed method easily handles the scalability by maintaining a good balance between the speed and the accuracy.

The utility based itemset mining approach has been discussed widely in recent years. There are many algorithms mining high utility itemsets by pruning candidates based on estimated utility values, and based on transaction-weighted utilization values. These algorithms aim to reduce search space. Besides, candidate pruning based on transaction-weighted utilization value is better than other strategies. In this paper, we propose TWU-Mining, a novel algorithm based-on WIT-tree for improving the cost of time and search space. Experiments show that the proposed algorithm is more effective on the testing databases.

In this paper, we propose a novel scheme called a self-verifying visual secret sharing scheme, which can be applied to both grayscale and color images. This scheme uses two halftone images. The first, considered to be the host image, is created by directly applying a halftoning technique to the original secret image. The other, regarded as the logo, is generated from the host image by exploiting the interpolation and error diffusion techniques. Because the set of shadows and the reconstructed secret image are generated by simple Boolean operations, no computational complexity and no pixel expansion occur in our scheme. Experimental results confirm that each shadow generated by our scheme is a noise-like image and eight times smaller than the secret image. Moreover, the peak signal-to-noise ratio value of the reconstructed secret image is larger than 33 dB. Based on the extracted halftone logo, the proposed scheme provides an effective solution for verifying the reliability of the set of collected shadows as well as the reconstructed secret image. Furthermore, the reconstructed secret image can be established completely if and only ifkout ofnvalid shadows have been collected. To achieve our objectives, four techniques were adopted: error diffusion, image clustering, interpolation, and inverse halftoning-based edge detection.

In this research, on the background of a combination between Neural Network and Fuzzy Logic theories applying to the field of recognition, we have developed our four-layer feedforward Fuzzy Neural Network (FNN) as a solution to solve Vietnamese character recognition problems. Our first-step experiments of FNN were applied for 28 Vietnamese accent characters in VNI-Times font. Our obtained testing results prove the availability of FNN as a good solution even to bad-quality patterns with high distortion rate.