This study introduces a novel encoding scheme for DNA/RNA sequences, integrating Komlós and Hadamard transforms. Unlike traditional One-Hot encoding, this approach offers a more informative representation of omics data while significantly reducing computational complexity. However, it is important to note that the Komlós transform component provides fewer features and does not utilize sparse codes. By leveraging the inherent properties of these transforms, our method effectively captures complex patterns within the data, leading to improved model accuracy and reduced training times. When combined with an image transformation, this encoding scheme demonstrates particularly efficient results, achieving superior performance across various predictive tasks with significantly lower computational resource demands compared to One-Hot encoding. Our findings suggest that this novel encoding scheme, particularly when integrated with Hilbert Curve mapping or sequence to image analysis, holds significant promise for advancing DNA/RNA data analysis by offering a more efficient and effective approach to feature representation.

The Industrial Internet of Things (IIoT) is increasingly vulnerable to sophisticated cyber-attacks, including zero-day and device-specific threats, which pose severe operational and security risks. We present a Federated Digital Twin for Spatio-Temporal Anomaly Detection approach that is tailored to offer a real-time, scalable and privacy-preserving security for IIoT settings. The framework integrates digital twin modeling to represent the physical network, spatio-temporal deep learning for capturing complex dependencies across devices and time, federated learning to enable collaborative model training without exposing sensitive data, and post-quantum cryptographic (PQC) security to ensure robust protection against emerging quantum-era threats. We evaluate the framework on the RT-IoT2022 dataset, demonstrating an overall detection accuracy of 99.89%, precision of 99.91%, recall of 99.87%, and an AUC of 0.998. The proposed method surpasses contemporary intrusion detection systems, accurately detects zero-day attacks, and reduces the overhead of communication by 65% in comparison with centralized training. Ablation studies demonstrate the importance of multi-view graph learning, cross-graph attention, and federated aggregation for obtaining good performance. Additionally, the framework offers transparent descriptions of the anomalies, so that operators can name and take response action against malicious operations. Its federated architecture ensures sensitive IIoT data remains on local devices, enhancing privacy and compliance with data protection regulations. The integration of PQC mechanisms future-proofs the system against quantum-era attacks, making it resilient to next-generation threats. Overall, the proposed framework offers a holistic, robust, and scalable solution for securing industrial networks and supports deployment across diverse IIoT infrastructures with minimal operational overhead.

EEG hyperscanning employs electroencephalography (EEG) activity to simultaneously monitor the brain activities of several individuals as they interact. During hyperscanning research, scientists pay attention to how brain activities of two or more subject become synchronized while doing a task. However, technological limitations associated with EEG hyperscanning-based brain-computer interfaces (BCIs) slowed down the exploration of this research for rehabilitation purposes. The method HyperCSP aims to minimize the impact of irrelevant actions that participants might perform naturally or deliberately. It effectively isolates a common motor task shared among several individuals. Recognizing this challenge, we developed HyperGCN, which identifies connectivity between the subjects' channels during the multi-common task. HyperGCN utilizes the principles of graph theory to analyze complex networks of brain activity. This allows a sophisticated interpretation of intra-brain connectivity. The HyperGCN helps to understand the patterns and interactions that traditional analysis methods miss by treating these connections as graphs. This method achieved an average resulting accuracy of 92.86% on the hyperscanning dataset. All code is publicly available on github11https://github.com/abduloo7454/XAI-Hyper-GCN/

Automated diabetic retinopathy (DR) grading is crucial for disease monitoring and personalized treatment, challenged by high intra-class variation and data imbalance. This paper presents a novel approach to enhancing DR grading detection by integrating linear-Radon sinogram-based images with original retinal images to provide a multimodal network using different convolutional neural network (CNN) architectures. Using the Kaggle Aptos dataset, we evaluated the performance of this multimodal integration. Our findings reveal a significant improvement in multi-class classification performance compared to unimodal retina-only images, underscoring our method's ability to detect subtle patterns among different DR grades. This study underscores the potential of sinogram-based images as a valuable modality for DR grading and paves the way for future research to validate this approach across diverse datasets and explore the application of curve-based sinograms,

Our study addresses the challenge of imbalanced regression data in Machine Learning (ML) by introducing tailored methods for different data structures. We adapt K-Nearest Neighbor Oversampling-Regression (KNNOR-Reg), originally for imbalanced classification, to address imbalanced regression in low population datasets, evolving to KNNOR-Deep Regression (KNNOR-DeepReg) for high-population datasets. For tabular data, we also present the Auto-Inflater neural network, utilizing an exponential loss function for Autoencoders. For image datasets, we employ Multi-Level Autoencoders, consisting of Convolutional and Fully Connected Autoencoders. For such high-dimension data our approach outperforms the Synthetic Minority Oversampling Technique for Regression (SMOTER) algorithm for the IMDB-WIKI and AgeDB image datasets. For tabular data we conducted a comprehensive experiment using various models trained on both augmented and non-augmented datasets, followed by performance comparisons on test data. The outcomes revealed a positive impact of data augmentation, with a success rate of 83.75% for Light Gradient Boosting Method (LightGBM) and 71.57% for the 18 other regressors employed in the study. This success rate is determined by the frequency of instances where models performed better when augmented data was used compared to instances with no augmentation. Access to the comparative code can be found in GitHub.

The detection and classification of neurological and psychological phenomena heavily rely on Electroencephalography (EEG). This study investigates the effectiveness of various feature extraction techniques and machine learning classifiers in EEG-based classification tasks. Stress detection using the Bird et al. dataset, which encompasses multiple emotional states, and seizure detection using the CHB-MIT dataset, known for its challenges in distinguishing seizure from non-seizure patterns, are specifically explored.The results highlight the crucial role of feature extraction methods in EEG-based classification. Among the techniques tested, our Progressive Fourier Transform (PFT) method consistently outperforms others, emerging as the superior choice.In stress detection, our proposed PFT achieves an outstanding accuracy of 98.41% on the Bird et al. dataset, surpassing existing methods based on statistical features. For seizure detection, our model attains a competitive accuracy of 96.88% on the CHBMIT dataset, showcasing efficiency even with a reduced number of channels.This study demonstrates the potential of EEG-based classification techniques in practical applications such as stress monitoring and seizure prediction. Furthermore, it emphasizes the significance of advanced feature extraction methods in achieving accurate results. Future research may involve refining these techniques further and expanding their applicability to diverse EEG datasets and other neurological and psychological disorders.

This paper describes the experimental framework and results of the IJCB 2022 Mobile Behavioral Biometrics Competition (MobileB2C). The aim of MobileB2C is bench-marking mobile user authentication systems based on behavioral biometric traits transparently acquired by mobile devices during ordinary Human-Computer Interaction (HCI), using a novel public database, BehavePassDB11https://github.com/BiDAlab/MobileB2C_BehavePassDE, and a standard experimental protocol. The competition is divided into four tasks corresponding to typical user activities: keystroke, text reading, gallery swiping, and tapping. The data are composed of touchscreen data and several background sensor data simultaneously acquired. “Random” (different users with different devices) and “skilled” (different user on the same device attempting to imitate the legitimate one) impostor scenarios are considered. The results achieved by the participants show the feasibility of user authentication through behavioral biometrics, although this proves to be a non-trivial challenge. MobileB2C will be established as an on-going competition22https://sites.google.com/view/mobileb2c/.

In this paper, a novel non-parametric clustering algorithm which is based on the concept of divide-and-merge is proposed. The proposed algorithm is based on two primary phases, after data cleaning: (i) the Division phase and (ii) the Merging phase. In the initial phase of division, the data is divided into an optimized number of small sub-clusters utilizing all the dimensions of the data. In the second phase of merging, the small sub-clusters obtained as a result of division are merged according to an advanced statistical metric to form the actual clusters in the data. The proposed algorithm has the following merits: (i) ability to discover both convex and non-convex shaped clusters, (ii) ability to discover clusters different in densities, (iii) ability to detect and remove outliers/noise in the data (iv) easily tunable or fixed hyperparameters (v) and its usability for high dimensional data. The proposed algorithm is extensively tested on 20 benchmark datasets including both, the synthetic and the real datasets and is found better/competing to the existing state-of-the-art parametric and non-parametric clustering algorithms.

This paper presents a chopper instrumentation amplifier design that employs a proposed amplifier slicing technique for offset reduction. In this scheme, the core amplifier is split into multiple identical slices. During operation, the offset polarity of these slices is firstly determined by employing the second-stage of the amplifier as a static comparator. Next, by using the polarity information, the amplifier slices are regrouped to achieve statistical offset suppression. A mathematical model is developed in this paper to estimate the effectiveness of this reduction scheme. The sliced amplifier structure also enables a scalable noise and bandwidth without adding extra analog components. Simulation results show that the proposed reduction scheme achieves a > 40 dB offset suppression and a noise efficiency factor (NEF) of 2.2. The circuit is implemented in a 0.18 μm standard CMOS technology for proof of concept and consumes 0.4 μA to 1 μA current from a 1.2 V supply to reach a noise level from 90 nV/√Hz to 31.8 nV/√Hz, respectively.

Machine learning is being implemented in bioinformatics and computational biology to solve challenging problems emerged in the analysis and modeling of biological data such as DNA, RNA, and protein. The major problems in classifying protein sequences into existing families/superfamilies are the following: the selection of a suitable sequence encoding method, the extraction of an optimized subset of features that possesses significant discriminatory information, and the adaptation of an appropriate learning algorithm that classifies protein sequences with higher classification accuracy. The accurate classification of protein sequence would be helpful in determining the structure and function of novel protein sequences. In this article, we have proposed a distance‐based sequence encoding algorithm that captures the sequence's statistical characteristics along with amino acids sequence order information. A statistical metric‐based feature selection algorithm is then adopted to identify the reduced set of features to represent the original feature space. The performance of the proposed technique is validated using some of the best performing classifiers implemented previously for protein sequence classification. An average classification accuracy of 92% was achieved on the yeast protein sequence data set downloaded from the benchmark UniProtKB database.

Classification of G protein-coupled receptors (GPCRs) according to their functions is an ongoing area of research which is helpful for the pharmaceutical industry in the development of drug targets for major diseases. Currently, more than 40% drugs in the market target GPCRs. The experimental methods of determining their function are very expensive and time consuming. Due to a rapid and constant increase in the GPCRs proteins in the public databases, it is extremely important to develop computational techniques that lessen the gap between the sequenced proteins and proteins with known functions. In this paper, a statistical method was utilized to encode proteins sequences. The encoding technique considers various distances for an amino acid in a sequence at different levels of decompositions. The Neural Network and Support Vector Machines classifiers were compared on 2 well-known GPCRs datasets. The results showed that better performance is achieved using neural network classifier. The classification accuracies were in the range of 94 to 98%.

Wireless Sensor Networks (WSNs) have been achieved widespread applicability in water quality monitoring. However, existing WSN-based monitoring systems are not adequate for monitoring pond and lake water, city water distribution and water reservoir. Moreover, these frameworks cannot be reused in other monitoring applications since they use static and application specific sensor nodes and are not dynamic to the changing requirements. Thus, we introduce a reusable, self-configurable, and energy efficient WSN-based water quality monitoring system that integrates a Web-based information portal and a sleep scheduling mechanism of sensor nodes. The testbed and simulation results show that the framework can monitor the water quality in real-time and the sleep scheduling mechanism increases the network lifetime, respectively.