Tri-perspective Multi-view Classification

Strategic bidding problems in energy markets have been vastly studied in the literature due to their economic importance and mathematical difficulty. In this paper, we consider this problem under the natural uncertainty associated to the competitors' bids and the system demand, and formulate it as a nonconvex quadratically constrained quadratic program. In order to solve this nonconvex maximization problem or to evaluate the quality of heuristic solutions, it is important to know good upper bounds to its optimal value. With this goal, we apply a cutting plane algorithm (CPA) to solve an extended linear relaxation of the problem, which is an outer approximation of a semidefinite relaxation. Valid inequalities are added to an initial linear relaxation at each iteration of the CPA improving the bound computed, including inequalities derived from reformulation linearization techniques (RLTs) and semidefinite programing (SDP) cuts that enforce the positive semidefiniteness of the matrix variable. SDP cuts are well known to considerably delay the solution of the relaxation because of their density, and therefore, we apply a sparsification procedure for the cuts, introduced by Margot et al. in . We propose a modification to the algorithm in Margot et al. to better control the density of the SDP cuts; and after tuning the parameters of our algorithm, we show the trade‐off between the bounds and the density of the cuts for test instances derived from the Brazilian electric system.

In a recent article, Bertsimas and Popescu showed that a tight upper bound on a Europeantype call option price, given the first n moments of the distribution of the underlying security price, can be obtained by solving an associated semidefinite programming problem (SDP). The purpose of this paper is to improve and extend their results. We will show that a tight lower bound can be calculated by solving another SDP. Also, we will show that these problems can be solved very quickly by a newly developed cutting plane algorithm when n is less than six or seven.

In this paper, we propose new approaches for action and event recognition by leveraging a large number of freely available Web videos (e.g., from Flickr video search engine) and Web images (e.g., from Bing and Google image search engines). We address this problem by formulating it as a new multi-domain adaptation problem, in which heterogeneous Web sources are provided. Specifically, we are given different types of visual features (e.g., the DeCAF features from Bing/Google images and the trajectory-based features from Flickr videos) from heterogeneous source domains and all types of visual features from the target domain. Considering the target domain is more relevant to some source domains, we propose a new approach named multi-domain adaptation with heterogeneous sources (MDA-HS) to effectively make use of the heterogeneous sources. In MDA-HS, we simultaneously seek for the optimal weights of multiple source domains, infer the labels of target domain samples, and learn an optimal target classifier. Moreover, as textual descriptions are often available for both Web videos and images, we propose a novel approach called MDA-HS using privileged information (MDA-HS+) to effectively incorporate the valuable textual information into our MDA-HS method, based on the recent learning using privileged information paradigm. MDA-HS+ can be further extended by using a new elastic-net-like regularization. We solve our MDA-HS and MDA-HS+ methods by using the cutting-plane algorithm, in which a multiple kernel learning problem is derived and solved. Extensive experiments on three benchmark data sets demonstrate that our proposed approaches are effective for action and event recognition without requiring any labeled samples from the target domain.

Image clustering is a classic problem in computer vision, which categorizes images into different groups. Recent studies utilize nouns as external semantic knowledge to improve clustering performance. However, these methods often overlook the inherent ambiguity of nouns, which can distort semantic representations and degrade clustering quality. To address this issue, we propose a hierarChical semAntic alignmEnt method for image clustering, dubbed CAE, which improves clustering performance in a training-free manner. In our approach, we incorporate two complementary types of textual semantics: caption-level descriptions, which convey fine-grained attributes of image content, and noun-level concepts, which represent high-level object categories. We first select relevant nouns from WordNet and descriptions from caption datasets to construct a semantic space aligned with image features. Then, we design a residual attention mechanism to further enhance the discriminability of this space. Finally, we combine the enhanced semantic and image features to perform clustering. Extensive experiments across 8 datasets demonstrate the effectiveness of our method, notably surpassing the state-of-the-art training-free approach with a 4.2% improvement in accuracy and a 2.9% improvement in adjusted rand index (ARI) on the ImageNet-1K dataset.

This paper presents a simple but powerful extension of the maximum margin clustering (MMC) algorithm that optimizes multivariate performance measure specifically defined for clustering, including Normalized Mutual In- formation, Rand Index and F-measure. Different from previous MMC algorithms that always employ the error rate as the loss function, our formulation involves a multivariate loss function that is a non-linear combination of the individual clustering results. Computationally, we propose a cutting plane algorithm to approximately solve the resulting optimization problem with a guaranteed accuracy. Experimental evaluations show clear improvements in clustering performance of our method over previous maximum margin clustering algorithms.

In this paper we present EPIC, an efficient and effective predictor for IC manufacturing hotspots in deep sub-wavelength lithography. EPIC proposes a unified framework to combine different hotspot detection methods together, such as machine learning and pattern matching, using mathematical programming/optimization. EPIC algorithm has been tested on a number of industry benchmarks under advanced manufacturing conditions. It demonstrates so far the best capability in selectively combining the desirable features of various hotspot detection methods (3.5-8.2% accuracy improvement) as well as significant suppression of the detection noise (e.g., 80% false-alarm reduction). These characteristics make EPIC very suitable for conducting high performance physical verification and guiding efficient manufacturability friendly physical design.

This brand new textbook will fill a currently empty niche in microbiology textbooks and provide a current, comprehensive view of microbial diversity. A microbial diversity course is considered one of three core courses that every department should require of its majors, other two being microbial genetics (Molecular Genetics of Bacteria, 4e) and pathogenesis (Bacterial Pathogenesis, 3e). There are typically two ways to teach courses on microbial diversity. 1) Focus on metabolic diversity: which microbes are able to eat metals, which microbes tolerate extremely high or extremely low temperatures, etc. 2) Consider microbes systematically or phylogenetically, figuring out which group belongs with or next to this other group, what line of descent might be between organism X and organism Y, etc.--a more evolutionary approach. Brown teaches from second approach. Intended for senior undergraduate students majoring in microbiology or a related field such as genetics or biology, graduate students either in other programs, or incoming graduate students with non-microbiology backgrounds. * Includes a series of Questions for Thought at end of each chapter. * Offers a practical guide to molecular phylogenetic analysis, focusing on how to create and interpret phylogenetic trees, and an overview or survey of the Tree of Life. * Focuses on major phylogenetic groups of bacteria and archaea. * Presents a discussion of various conceptual aspects of microbial complexity, from genomics and cell biology. This title is published by American Society of Microbiology Press and distributed by Taylor and Francis in rest of world territories.

This study proposes a novel modification to the extreme learning machine (ELM) by introducing an optimized least-squares formulation with penalty regularization, enabling stable and efficient handling of sparse, high-dimensional multimedia data. In contrast to traditional autoencoders, our approach incorporates ELM-driven hidden layer refinement, achieving more effective feature compression and dimensionality reduction tailored for multimedia churn prediction. Additionally, we design a Gaussian kernel adaptation specifically for multimedia datasets, which replaces random feature mappings and enhances both predictive robustness and generalization performance. Empirical validation on a public multimedia customer behavior dataset reveals the model’s superiority over traditional churn prediction methods, showcasing notable improvements in prediction accuracy and precision. This research equips businesses with a robust model for informed CRM decision-making, enhancing customer retention and profitability. The study contributes to the in-depth understanding of multimedia data’s role in fostering sustainable and profitable customer relationships in today’s digital era.

Graph representation learning has made major strides over the past decade. However, in many relational domains, the input data are not suited for simple graph representations as the relationships between entities go beyond pairwise interactions. In such cases, the relationships in the data are better represented as hyperedges (set of entities) of a non-uniform hypergraph. While there have been works on principled methods for learning representations of nodes of a hypergraph, these approaches are limited in their applicability to tasks on non-uniform hypergraphs (hyperedges with different cardinalities). In this work, we exploit the incidence structure to develop a hypergraph neural network to learn provably expressive representations of variable sized hyperedges which preserve local-isomorphism in the line graph of the hypergraph, while also being invariant to permutations of its constituent vertices. Specifically, for a given vertex set, we propose frameworks for (1) hyperedge classification and (2) variable sized expansion of partially observed hyperedges which captures the higher order interactions among vertices and hyperedges. We evaluate performance on multiple real-world hypergraph datasets and demonstrate consistent, significant improvement in accuracy, over state-of-the-art models.

Diseases of the Gastrointestinal (GI) tract significantly affect the quality of human life and have a high fatality rate. Accurate diagnosis of GI diseases plays a pivotal role in healthcare systems. However, processing large amounts of medical image data can be challenging for radiologists and other medical professionals, increasing the risk of inaccurate medical assessments. Computer‐aided Diagnosis systems provide help to doctors for rapid and accurate diagnosis, thus resulting in saving lives. Recently, many techniques are found in the literature that uses deep Convolutional Neural Network (CNN) models for accurate disease classification. However, they have limitations in their ability to detect deformation‐invariant features and lack robustness. The diseased region is highlighted, using attention‐based image generation and superimposition with original images. A lightweight deep CNN model is employed to get significant features. These features are further reduced using a Cosine similarity‐based technique. The proposed framework is assessed using the Kvasir dataset. To verify the effectiveness of the proposed framework, vast experiments are conducted. The overall accuracy of 97.68%, 99.02% precision, 96.37% recall, and an F‐measure of 97.68% are achieved using the 810 significant features. This reduction in features resulted in a significant reduction in classification time. The robustness of the framework can be observed not only in terms of considerable improvement in accuracy, but also in terms of precision as well as recall, and F‐measure.

This paper presents two modified composite kernel frameworks for the hyperspectral image classification. The first proposed framework is a simple extension of the previously proposed generalized composite kernel (GCK) approach, and in the second framework, we have used a voting strategy. The developed modified composite kernels combine different spatial and spectral profiles without considering any weight parameter. We have used the multinomial logistic regression (MLR) algorithm for the classification of hyperspectral images in this work. We have developed different spatial profiles: extended morphological profile (EMP) and extended multiattribute profile (EMAP) in this work, and further integrated them with the spectral profile for spectral-spatial classification. Experimental results with the AVIRIS Indian Pines, ROSIS Pavia university area and AVIRIS Salinas images indicate that the proposed framework leads to an improvement in classification accuracy. The obtained overall accuracy increases by 2–3% for AVIRIS Indian Pines scene, 1–2% for ROSIS Pavia University Area scene and 1–3% for AVIRIS Salinas scene with the first proposed framework. In case of the second proposed framework with a sufficient number of training samples per class, the overall accuracy increases by 3–7% for AVIRIS Indian Pines scene, 2.5% for ROSIS Pavia University Area scene and 7–8% for AVIRIS Salinas scene. For this framework (Modified-CKV with MLR) we have obtained the overall accuracy of 96.95% for the Indian Pines scene, 99.33% for the Pavia University Area, and 97.76% for the Salinas scene with 75 training samples per class, which is less than 2% of the available labelled samples. With the results obtained, we have also concluded that with the linear combination approach for the construction of composite kernels, the overall accuracy improves without any increase in computational complexity.

In this paper, we discuss the intermediate results of our on-going study of DeepFakes detection in videos. Our core focus is in exploitation of feature engineering as a precursor filtering technique, to the deep learning-based convolution neural network (CNN) classification frameworks. In previous research, we focused on the standard deviation of the points of interest from SIFT or Scale-Invariant Feature Transform, to detect whether visual media has been compromised with misinformation. Hence, in this new approach we rely on classical frequency analysis of images that reveals different behaviors at higher frequencies. We noticed in literature review that there lies a distinct contrast of range of frequency component that emphasize important information in images [11]. We plan to use Discrete Wavelet Transform, abbreviated as DWT, and anticipate improvement of detection accuracy when used on complex and poor-quality images of FaceForensics++. The DWT features will be input to the CNN binary classification to further analyze the performance essentially when detecting this fraudulent information and to compare the detection performance against previous research which used Scale-Invariant Feature Transform.

Joint learning on diverse ultrasonic source scenarios presents a challenge in preserving stable generalization due to the combination of heterogeneity of different sources and the inconsistency of joint learning features. Previous joint learning studies, which are not source-resilient frameworks, may not preserve stable generalization when trained on diverse source scenarios. Furthermore, the limited variations in single-source data and the interference from ultrasound imaging, which are common in ultrasonic source scenarios, further decrease generalization. To address these problems, we proposed a source-resilient joint learning framework consisting of three stages: 1) Source transforming, where our 1-to-N transformation unifies diverse source scenarios for source-resiliency. 2) Our feature enhancement modules model the source-resilient joint learning network, including a manifold-constraint normalization module (MCNM) for addressing heterogeneity by minimizing manifold-based loss, a task-consistent attention module (TCAM) shares the multi-scale features with self-attention to address inconsistency, and an adaptive feature-shifting module (AFSM) for feature-level augmentation to overcome single-source data. 3) Our ultrasound-hybrid linear mapping (USmapping) cascades speckle randomization and mask-guiding Monge-Kantorovitch linear mapping to achieve ultrasonic style randomization for addressing the interference of ultrasonic data. Our framework was evaluated on eight ultrasound datasets from various scanners at multiple centers and surpassed previous comparable studies in both segmentation (DSC$WAvg$ of 75.7%) and classification (AUROC$WAvg$ of 68.8%) tasks. Our framework has the potential to serve as a general framework for enhancing the performance of joint learning under diverse ultrasonic source scenarios.

Hand fractures, particularly phalangeal fractures, are often subtle, concealed, and complex, posing significant challenges for accurate and efficient diagnosis with traditional radiography. To address the limitations of existing deep learning methods in handling complex anatomy, multi‐view uncertainty, and causal discrimination, we propose a novel two‐stage context‐aware multi‐view fusion framework for precise fracture classification and localization. In the first stage, a Context‐Aware Multi‐View Classification Network (CAMVC‐Net) is developed. A Dual‐path Semantic‐Spatial Alignment (DSSA) module aligns low‐level spatial details with high‐level semantics, enhancing fine‐grained fracture discrimination. A Context Aware Pyramid (CAP) module captures multi‐scale context, while a Counterfactual Attention Learning (CAL) loss guides the network to focus on discriminative regions. Multi‐view uncertainty is modeled using a Dirichlet distribution, and decision‐level fusion is achieved by Dempster's rule to mimic clinical reasoning. In the second stage, a Detail‐aware Fracture Localization Network (DFL‐Net) is designed. To adapt to irregular fracture geometry, DFL‐Net integrates deformable convolutions and incorporates DSSA into the feature pyramid to preserve fine spatial details during deep downsampling. Experiments on the MURA and clinical datasets demonstrate strong performance: the classification model achieved an accuracy of 92.89%, precision of 93.05%, recall of 96.96%, and F1 score of 94.97%. The localization network obtained 73.8% AP 50 and 43.8% AP 75 , with a 10.4% improvement in AP 75 over Faster R‐CNN. These results indicate that the proposed framework provides an accurate and efficient tool for computer‐aided fracture diagnosis, with the potential to reduce misdiagnosis and improve clinical decision‐making efficiency.

This study investigates a late fusion framework for breast cancer classification in mammography using deep learning. Two EfficientNet-B0 models were trained independently on craniocaudal (CC) and mediolateral oblique (MLO) mammographic views. Their outputs were integrated via a fully connected late fusion layer. Experimental evaluation on a dataset collected from two hospitals showed that singleview models achieved 82.14 % accuracy with perfect sensitivity but limited specificity, while the late fusion model improved performance to 92.86 % accuracy, 1.000 sensitivity, and 0.8667 specificity. The proposed lightweight fusion framework is computationally efficient and particularly suitable for small, clinically curated datasets. These results confirm that multiview late fusion reduces false positives and enhances diagnostic reliability in computer-aided mammography. Moreover, our deterministic and lightweight fusion enables faster inference compared with probabilistic or transformer-based fusion methods.