Hyperspectral imaging (HSI) has proven to be a powerful tool for capturing detailed spectral and spatial information across diverse applications. Despite the advancements in deep learning (DL) and Transformer architectures for HSI classification, challenges such as computational efficiency and the need for extensive labeled data persist. This letter introduces WaveMamba, a novel approach that integrates wavelet transformation with the spatial-spectral Mamba (SSMamba) architecture to enhance HSI classification. WaveMamba captures both local texture patterns and global contextual relationships in an end-to-end trainable model. The Wavelet-based enhanced features are then processed through the state-space architecture to model spatial-spectral relationships and temporal dependencies. The experimental results indicate that WaveMamba surpasses existing models, achieving an accuracy improvement of 4.5% on the University of Houston dataset and a 2.0% increase on the Pavia University dataset.

Hyperspectral imaging provides rich spectral-spatial information essential for fine-grained land cover classification. However, high dimensionality, spectral redundancy, and noise sensitivity significantly hinder classification accuracy. To overcome these issues, this work propose DIFF-SE, a novel differential transformer framework enhanced with a dual-path squeeze-and-excitation (E-SE) module tailored for hyperspectral image (HSI) classification (HSIC). The proposed multi-head differential attention mechanism contrasts paired attention maps to amplify discriminative spectral-spatial cues while suppressing redundancy and noise. Simultaneously, the E-SE module performs concurrent spectral and spatial recalibration, dynamically emphasizing informative bands and salient regions. Extensive experiments on three benchmark datasets, Pavia University (PU), WHU-Hi-HanChuan (HC), and OHID-1, demonstrate that DIFF-SE consistently achieves superior overall accuracies of 99.34%, 99.31%, and 94.99%, respectively, outperforming several recent state-of-the-art methods. The source code will be publicly released at https://github.com/mahmad000.

Hyperspectral images (HSIs) capture detailed spectral–spatial information across hundreds of contiguous bands, enabling precise material identification in domains such as environmental monitoring, agriculture, and urban analysis. However, the high dimensionality and spectral variability inherent to HSIs present significant challenges for effective feature extraction and classification. This letter introduces EnergyFormer (EF), a transformer-based framework designed to overcome these limitations through three key innovations: 1) multihead energy attention (MHEA), which formulates an energy optimization mechanism to selectively enhance discriminative spectral–spatial features; 2) Fourier positional embedding (FoPE), which adaptively models long-range spectral and spatial dependencies; and 3) enhanced convolutional block attention module (ECBAM), which emphasizes informative wavelength bands and spatial structures for robust representation learning. Extensive experiments on the WHU-Hi-HanChuan, Salinas, and Pavia University datasets demonstrate that EF achieves superior classification performance with overall accuracies of 99.28%, 98.63%, and 98.72%, respectively, outperforming leading CNN-, transformer-, and Mamba-based models.

Hyperspectral image classification (HSIC) is a challenging task due to the high dimensionality of hyperspectral data, the complex interplay of spatial and spectral features, and the scarcity of annotated samples. Existing approaches, mainly based on tokenization-based feature extraction, introduce artificial segmentation, increasing computational cost, and may lead to information loss. To address these issues, a novel framework, Byte Latent Mamba with Knowledge Distillation (BLM-KD), overcoming explicit tokenization by directly learning byte-level spectral-spatial representations from raw hyperspectral data, is proposed. The Byte Latent Mamba architecture learns compact and expressive byte-level features through an end-to-end convolutional encoder, preserving spectral continuity and spatial structure. A structured State Space Model (SSM) is integrated to model long-range spatial-spectral dependencies efficiently via learned dynamic state transitions. Additionally, an adaptive knowledge distillation (KD) strategy is adopted, where a high-capacity teacher model selectively transfers salient features to a lightweight student model, driven by a temperature-controlled weighting schedule. This ensures robust generalization with reduced model complexity. A patch-based preprocessing scheme also excludes irrelevant zero-labeled samples, refining the training process. Extensive experiments conducted on multiple real-world hyperspectral benchmarks demonstrate that BLM-KD outperforms existing state-of-the-art methods in both classification accuracy and computational efficiency.

Multihead self-attention and cross-attention mechanisms often suffer from computational inefficiencies, limited scalability, and suboptimal contextual understanding, particularly in hyperspectral image (HSI) classification. These mechanisms struggle to effectively capture long-range dependencies while maintaining computational feasibility due to the quadratic complexity of self-attention. To address these challenges, this work proposes PolicyMamba, a spectral-spatial mamba model enhanced with a localized policy attention mechanism. This mechanism reduces computational overhead by restricting attention to nonoverlapping localized regions and enforcing sparsity constraints, ensuring that only the most informative interactions are retained. A hierarchical aggregation strategy further integrates patch-wise attention outputs, preserving spectral-spatial correlations across scales. In addition, a sliding window patch process enhances local feature continuity while mitigating information loss. The PolicyMamba framework integrates spectral-spatial token generation, token enhancement, localized attention, and state transition modules, significantly improving HSI feature representation. Extensive experiments demonstrate that PolicyMamba achieves superior classification accuracy, outperforming conventional and state-of-the-art methods in land cover classification (LCC) by efficiently modeling intricate dependencies in HSI data.

Conventional active learning approaches for hyperspectral image classification (HSIC) have limitations such as incrementally growing training sets without considering class structure and heterogeneity within existing and new samples. Additionally, there is limited research leveraging both spectral and spatial information jointly, and stopping criteria are not well established. This study presents a novel fuzzybased spatial-spectral Within and Between method (FLG) for preserving local and global class discriminative information. The method first explores spatial fuzziness to identify misclassified samples. It then computes total within-class and between-class information locally and globally. This information is integrated into a discriminative objective function to selectively query heterogeneous samples, mitigating randomness among training data. Experimental results on benchmark Hyperspectral datasets demonstrate the FLG improves classification accuracy across generative, extreme learning machine, and sparse multinomial logistic regression models by jointly exploiting spectral and spatial information to expand labeled training sets strategically.

In this paper, we address the challenge of learning to recite the Quran for non-Arabic speakers. We explore the possibility of crowdsourcing a carefully annotated Quranic dataset, on top of which AI models can be built to simplify the learning process. In particular, we use the volunteer-based crowdsourcing genre and implement a crowdsourcing API to gather audio assets. We integrated the API into an existing mobile application called NamazApp for collecting audio recitations. We developed a crowd-sourcing platform called Quran Voice for annotating the gathered audio assets. As a result, we have collected around 7000 Quranic recitations from a pool of 1287 participants across more than 11 non-Arabic countries, and we have annotated 1166 recitations from the dataset in six categories. We have achieved a crowd accuracy of 0.77, an inter-rater agreement of 0.63 between the annotators, and 0.89 between the labels assigned by the algorithm and the expert judgments. 1

Quantum computing (QC) has become a fascinating and popular topic due to its broad range of applications, particularly in machine learning (ML). The intersection of QC and ML is known as Quantum Machine Learning (QML). QML is a field that investigates how QC can enhance ML, making it one of the most exciting areas of research due to the potential of QC in solving complex problems. In this paper, we demonstrate the Quantum Annealing (QA) approach to improving ML in binary classification tasks. We implemented the QBoost algorithm on a few datasets using D-Wave’s quantum computers, specifically the Advantage 1 and Advantage 2 prototypes, incorporating the new feature Fast Anneal.

Transformers have proven effective for Hyperspectral Image Classification (HSIC) but often incorporate average pooling that results in information loss. This paper presents WaveFormer, a novel transformer-based approach that leverages wavelet transforms for invertible downsampling. This preserves data integrity while enabling attention learning. Specifically, WaveFormer unifies downsampling with wavelet transforms to decompress feature maps without loss. This provides an efficient tradeoff between performance and computation. Furthermore, the wavelet decomposition enhances the interaction between structural and shape information in image patches and channel maps. To evaluate WaveFormer, we conducted extensive experiments on two benchmark hyperspectral datasets. Our results demonstrate that WaveFormer achieves state-of-the-art classification accuracy, obtaining overall accuracies of 95.66% and 96.54% on the Pavia University and the University of Houston datasets, respectively. By integrating wavelet transforms, WaveFormer presents a new transformer architecture for hyperspectral imagery that achieves superior classification without information loss from average pooling.

Hyperspectral image classification (HSIC) faces challenges in preserving high-frequency features during downsampling and hierarchical filtering in the CNN architecture. To overcome this, we propose sharpened cosine similarity (SCS) as an alternative to convolutions within a neural network for HSIC. SCSNet emphasizes parameter efficiency by bypassing nonlinear activation layers, normalization steps, and dropout post the SCS layer. Additionally, MaxAbsPool is implemented instead of MaxPool for superior performance. Experimental results on public HSI datasets demonstrate SCS’s comparable accuracy, achieving 99% for both Indian Pines and Salinas datasets.

In Transformer-based hyperspectral image classification (HSIC), predefined positional encodings (PEs) are crucial for capturing the order of each input token. However, their typical representation as fixed-dimensional learnable vectors makes it challenging to adapt to variable-length input sequences, thereby limiting the broader application of Transformers for HSIC. To address this issue, this study introduces an implicit conditional PEs (CPEs) scheme in a Transformer for HSIC, conditioned on the input token’s local neighborhood. The proposed spatial–spectral Transformer (SSFormer) integrates spatial–spectral information and enhances classification performance by incorporating a CPE mechanism, thereby increasing the Transformer layers’ capacity to preserve contextual relationships within the HSI data. Moreover, SSFormer ensembles the cross attention between patches and proposed learnable embeddings. This enables the model to capture global and local features simultaneously while addressing the constraint of limited training samples in a computationally efficient manner. Extensive experiments on publicly available HSI benchmarking datasets were conducted to validate the effectiveness of the proposed SSFormer model. The results demonstrated remarkable performance, achieving the classification accuracies of 97.7% on the Indian Pines dataset and 96.08% on the University of Houston dataset.

Hyperspectral images (HSIs) are used in a large number of real-world applications. HSI classification (HSIC) is a challenging task due to high interclass similarity, high intraclass variability, overlapping, and nested regions. The 2-D convolutional neural network (CNN) is a viable classification approach since HSIC depends on both spectral–spatial information. The 3-D CNN is a good alternative for improving the accuracy of HSIC, but it can be computationally intensive due to the volume and spectral dimensions of HSI. Furthermore, these models may fail to extract quality feature maps and underperform over the regions having similar textures. This work proposes a 3-D CNN model that utilizes both spatial–spectral feature maps to improve the performance of HSIC. For this purpose, the HSI cube is first divided into small overlapping 3-D patches, which are processed to generate 3-D feature maps using a 3-D kernel function over multiple contiguous bands of the spectral information in a computationally efficient way. In brief, our end-to-end trained model requires fewer parameters to significantly reduce the convergence time while providing better accuracy than existing models. The results are further compared with several state-of-the-art 2-D/3-D CNN models, demonstrating remarkable performance both in terms of accuracy and computational time.

Present-day path planning algorithms for UAVs rely on various parameters that need to be tuned at runtime to be able to plan the best possible route. For example, for a sampling-based algorithm, the number of samples plays a crucial role. The dimension of the space that is being searched to plan the path, the minimum distance for extending a path in a direction, and the minimum distance that the drone should maintain with respect to obstacles while traversing the planned path are all important variables. Along with this, we have a choice of vision algorithms, their parameters, and platforms. Finding a suitable configuration for all these parameters at runtime is very challenging because we need to solve a complicated optimization problem, and that too within tens of milliseconds. The area of theoretical exploration of the optimization problems that arise in such settings is dominated by traditional approaches that use regular nonlinear optimization techniques often enhanced with AI-based techniques such as genetic algorithms. These techniques are sadly rather slow, have convergence issues, and are typically not suitable for use at runtime. In this article, we leverage recent and promising research results that propose to solve complex optimization problems by converting them into approximately equivalent game-theoretic problems. The computed equilibrium strategies can then be mapped to the optimal values of the tunable parameters. With simulation studies in virtual worlds, we show that our solutions are 5-21% better than those produced by traditional methods, and our approach is 10× faster.

Modern transportation systems, such as computer networks, have become increasingly faster, aiming to “shorten” distances and travel time. This trend allows thinking about new services and induces to reconsider existing ones starting from new technologies, as for Intelligent Transportation Systems. Thereby, an all-encompassing scenario laying at the intersection of several domains, including manufacturing, logistics, traveling, insurance, maintenance, and trading, with the transportation one, can be envisioned. The building block for the resulting transportation ecosystem is an information system that is able to gather and connect all involved stakeholders and domains around the concept of mobility and vehicle to address complex multifaceted problems in an efficient and trustworthy way. This paper proposes the adoption of distributed ledgers to implement such a vehicle-centric information system, distributing data across the network while ensuring trustworthiness. Starting from the vehicle lifecycle immutable and certified information, new services for cross-cuttingly addressing diversity and complexity in the transportation ecosystem can be implemented. The proposed solution merges Multichain and MongoDB technologies to achieve a trade-off between trustworthiness and performance, storing only the metadata in the Multichain network. Its effectiveness is demonstrated by an example on a vehicle trading service showing an overhead for the proposed solution below 18% against a pure MongoDB one on I/O operations.

An increasing interest is growing around the idea of microservices and the promise of improving scalability when compared to monolithic systems. Several companies are evaluating pros and cons of a complex migration. In particular, financial institutions are positioned in a difficult situation due to the economic climate and the appearance of agile competitors that can navigate in a more flexible legal framework and started their business since day one with more agile architectures and without being bounded to outdated technological standard. In this paper, we present a real world case study in order to demonstrate how scalability is positively affected by re-implementing a monolithic architecture (MA) into a microservices architecture (MSA). The case study is based on theFX Coresystem, a mission critical system of Danske Bank, the largest bank in Denmark and one of the leading financial institutions in Northern Europe. The technical problem that has been addressed and solved in this paper is the identification of a repeatable migration process that can be used to convert a real world Monolithic architecture into a Microservices architecture in the specific setting of financial domain, typically characterized by legacy systems and batch-based processing on heterogeneous data sources.

The Internet of Things (IoT) and Robotics cannot be considered two separate domains these days. The Internet of Robotics Things (IoRT) is a concept that has been recently introduced to describe the integration of robotics technologies in IoT scenarios. As a consequence, these two research fields have started interacting, and thus linking research communities. In this paper we intend to make further steps in converging the two communities and broaden the discussion on the development of this interdisciplinary field. The paper provides overview, analysis and challenges of possible solutions for the Internet of Robotic Things, discussing the issues of the IoRT architecture, and the integration of smart environments and robotic applications.

Social media are a way for people to build their reputation or to promote an idea. Twitter, in contrast with other social media sources, is a generator of real-time textual information, and it is mainly used to share ideas, opinions and breaking news. It is meant for short, quick, compelling statements that reach out millions of users around the world. Posting something inappropriate may affect the public image, privacy of celebrities, politicians as well as ordinary Twitter users. If we could in advance alarm the user of the potential vulnerability in the message to be posted we could protect his/her identity from being compromised. So, automatic identification of the message with the content causing it to be deleted in the future is a promising area of research. In this paper, we are analyzing Twitter messages in English language with the objective to build a classifier to predict whether a particular post will be deleted by the user or not. We apply the Recurrent Neural Networks (RNN) model that relies on the context-based information of tweets while doing the classification. An additional contribution of the work is the construction of a rich set of features including twitter metadata, user information and tweets' text to train classical machine learning algorithms on Twitter data.

The vision encompassing Smart and Software-defined Buildings (SSDB) is becoming more popular and its implementation is now more accessible due to the widespread adoption of the Internet of Things (IoT) infrastructure. Some of the most important applications sustaining this vision are energy management, environmental comfort, safety and surveillance. This paper surveys IoT and SSB technologies and their cooperation towards the realization of smart spaces. We propose a four-layer reference architecture and we organize related concepts around it. This conceptual frame is useful to identify the current literature on the topic and to connect the dots into a coherent vision of the future of residential and commercial buildings.

Catastrophic forgetting has a significant negative impact in reinforcement learning. The purpose of this study is to investigate how pseudorehearsal can change performance of an actor-critic agent with neural-network function approximation. We tested agent in a pole balancing task and compared different pseudorehearsal approaches. We have found that pseudorehearsal can assist learning and decrease forgetting.

In this paper, we investigate how dynamic properties of reputation can influence the quality of users' ranking. Reputation systems should be based on rules that can guarantee high level of trust and help identify unreliable units. To understand the effectiveness of dynamic properties in the evaluation of reputation, we propose our own model (DIB-RM) that utilizes three factors: forgetting, cumulative, and activity period. In order to evaluate the model, we use data from StackOverflow which also has its own reputation model. We estimate similarity of ratings between DIB-RM and the StackOverflow reputation model to test our hypothesis. We use two values to calculate our metrics: DIB-RM reputation and historical reputation. We found out that historical reputation gives better metric values. Our preliminary results are presented for different sets of values of the aforementioned factors in order to analyze how effectively the model can be used for modeling reputation systems.

Multiplayer computer games play a big role in the ever-growing entertainment industry. Being competitive in this industry means releasing the best possible software, and reliability is a key feature to win the market. Computer games are also actively used to simulate different robotic systems where reliability is even more important, and potentially critical. Traditional software testing approaches can check a subset of all the possible program executions, and they can never guarantee complete absence of errors in the source code. On the other hand, during more than twenty years, Model Checking has demonstrated to be a powerful instrument for formal verification of large hardware and software components. In this paper, we contribute with a novel approach to formally verify computer games. We propose a method of model construction that starts from a computer game description and utilizes Model Checking technique. We apply the method on a case study: the game Penguin Clash. Finally, an approach to game model reduction (and its implementation) is introduced in order to address the state explosion problem.

Microservices is an architectural style inspired by service-oriented computing that has recently started gainingpopularity. Jolie is a programming language based on the microservices paradigm: the main building block of Jolie systems are services, in contrast to, e.g., functions or objects. The primitives offered by the Jolie language elicit many of the recurring patterns found in microservices, like load balancers and structured processes. However, Jolie still lacks some useful constructs for dealing with message types and data manipulation that are present in service-oriented computing. In this paper, we focus on the possibility of expressing choices at the level of data types, a feature well represented in standards for Web Services, e.g., WSDL. We extend Jolie to support such type choices, and enable Jolie processes to act on data generically (without knowing which type it has in the choice). We show the impact of our implementation on some of the typical scenarios found in microservice systems. This shows how computation can move from a process-driven to a data-driven approach, and leads to the preliminary identification of recurring communication patterns that can be shaped as design patterns.

This paper introduces a new model of artificial cognitive architecture for intelligent systems, the Neuromodulating Cognitive Architecture (NEUCOGAR). The model is bio mimetically inspired and adapts the neuromodulators role of human brains into computational environments. This way we aim at achieving more efficient Artificial Intelligence solutions based on the biological inspiration of the deep functioning of human brain, which is highly emotional. The analysis of new data obtained from neurology, psychology philosophy and anthropology allows us to generate a mapping of monoamine neuro modulators and to apply it to computational system parameters. Artificial cognitive systems can then better perform complex tasks (regarding information selection and discrimination, attention, innovation, creativity) as well as engaging in affordable emotional relationships with human users.

To actually bridge the gap between business perspective and technical perspective, the prerequisite is to provide a comprehensive process modeling framework for business processes. Different from the previous traditional process methodologies, our work is neither only industrial process graphic-modeling nor pure theoretical studies. We mainly focus on the semantically-enhanced process description model. We propose the semantic modeling framework for business processes, i.e. BPMO: basically, we determine the description requirements for the whole business process lifecycle involving process discovery, composition and execution; furthermore, we refine the comprehensive semantic Web services conceptual model WSMO and make specific extensions to realize the BPMO modeling framework.