M. Priyadharshini

Multi-label classification (MLC) seeks to make multiple predictions for an instance by identifying associations between labels, thereby enhancing prediction capability. The paper presents a new method for combining improved chicken swarm optimization (ICSO) feature selection with recurrent neural networks (RNNs) for MLC. ICSO overcomes feature selection issues and minimizes noise and redundancy in imbalanced datasets, whereas RNNs learn label dependencies to achieve higher accuracy. It is initially normalized using the Z-score method and then analyzed for dimensionality reduction using principal component analysis (PCA). Our approach is more accurate, more precise, better at recall, and achieves higher F-measure and specificity, and a lower error rate than current methods such as multi-label k-nearest neighbors (ML-KNN), fuzzy rough set learning with label-specific features (FRS-LIFT), and adaptive synthetic data for multi-label classification (ASD-MLC). The paper shows that ICSO is useful for augmenting RNN-based multi-label classification and may be applied in medical diagnosis and bioinformatics.

Solving ordinary differential equations (ODEs) constitutes a fundamental problem for many scientific and engineering disciplines, particularly for stiff, high-dimensional problems, or problems with changing dynamics. Consequently, traditional numerical solvers, such as the Runge–Kutta methods, suffer from the drawback of computational inefficiency, numerical in stability and a lack of good generality across a range of problems. However, some of these limitations had been overcome with the advent of deep learning methods, leading to the development of neural ordinary differential equations (Neural ODEs) methods. However, the Neural ODEs fixed architecture makes it less versatile and especially in cases of solving complex and nonlinear ODEs problems. To address these problems, this research proposes self-evolving meta-learning neural ODE (SEML-NODE), a new approach of meta-learning and evolutionary computing for making a self-adjusting neural solver that is capable of evolving its own architecture during training. The main novelty of SEML-NODE is the self-evolution mechanism that enables the neural network's structure to expand and/or contract according to error feedback so that both the accuracy and generalization are improved. The adaptive training process renders the new extrapolation method to have superior performance on the stiff or chaotic ODEs. In this work, we evaluate SEML-NODE on benchmark problems, i.e. logistic growth model, Kepler's problem, and Fisher's equation, which contain difficulty such as nonlinearity, stiffness, and spatial-temporal dynamical problem. The results show that SEML-NODE is consistently better than state-of-the-art methods, such as fixed Neural ODEs and physics-informed neural networks, achieving significantly improved results in terms of errors on all benchmark tasks. Since the framework is able to dynamically change architecture depending on complexity of problem, by solving more complex, time-varying systems is very efficient, which is a problem for traditional solvers. While the current dynamic network expansion has a big initial cost in computed power, it becomes more efficient as the network training progresses by eliminating unnecessary network complexity and only expanding as needed. However, the improved optimization of self-evolution parameters (such as the error threshold of network expansion) has been challenging to prevent overfitting. To overcome this, for future work, the optimization of these parameters and the extension of the SEML-NODE to partial differential equations will be the focus. Besides, reinforcement learning will also be explored to further optimize the architecture. The proposed SEML-NODE framework greatly improves the adaptive ODE solver development and may find many applications in real-world areas such as climate prediction, drug discovery, and financial prediction, where the traditional numerical methods cannot provide efficient solutions.

Psoriasis is a skin disorder which mainly occurs as a rash, scaly areas and an itchy skin. The symptoms usually occur on the chest, elbows, and the scalp. World Health Organization (WHO) reports that about 125 million individuals in the world, that is about two to three% of the world population, live with psoriasis. Moreover, approximately 30% of psoriasis patients can have psoriatic arthritis. Psoriasis can have a detrimental effect on the quality of life of the affected people. They can also deal with several chronic diseases, such as depression, metabolic syndrome, cardiovascular diseases, such as atherosclerosis, heart attacks, and strokes. The psoriasis may develop at any age of life, but this disease is most prevalent in the ages between 20 and 30 and 50–60. Psoriasis is caused by different aspects, such as hereditary aspects, lifestyle habits of modern people, diet, use of drugs, skin trauma, stress, or abnormalities in the immune system, and hormonal fluctuations. Regrettably, the available conventional measures of determining the type of psoriasis are not always accurate with human errors. In order to overcome these problems, our study would attempt to develop a new dataset classified into seven classes of psoriasis diseases. We make use of publicly accessible data like SKIN LESION, ISIC and DEMANET. To overcome the problem of class imbalances, we will use the method of image augmentation to make the number of imageries in each class close to each other.To identify the presence of psoriasis in skin disease images, we employ transfer learning algorithms. Specifically, our proposed methodology utilizes ResNet50, InceptionResNetV2, and InceptionV3 with Adam and RMSprop optimizers for psoriasis classification. During training and validation, the ResNet50 model achieved good accuracy rates with values of 92.36, 84.59, and 83.55, respectively. The InceptionV2 model demonstrated impressive accuracy during training, validation, and testing, with values of 99.07, 96.65, and 97.20, respectively. Similarly, the InceptionV3 model achieved superior accuracy rates during training, validation, and testing, with values of 99.57, 96.82, and 98.68, respectively.When compared to all the models, InceptionV3 consistently demonstrates superior accuracy.

Ordinary differential equations (ODEs) are very basic when it comes to modeling dynamic systems in various fields of science and engineering. However, solving high‐dimensional, nonlinear, and stiff ODEs is still a major challenge given the limitations of existing numerical methods, which tend to have difficulties in terms of accuracy and efficiency. This paper presents a new numerical approach called the enhanced differential transform homotopy perturbation method (E‐DTHPM), dedicated to overcoming the above difficulties. E‐DTHPM presents a combination of the benefits of the differential transform method (DTM) and homotopy perturbation method (HPM) with an adaptive step sizing mechanism that is dynamic in step sizing based on local truncation errors (LTEs). The combination of these methods addresses an important research gap by offering an efficient and accurate solution to the IVP of ODEs with nonsmooth behaviors (singularities and discontinuity behaviors). The proposed method is an improvement over conventional techniques and will provide increased computational efficiency and precision for complicated high‐dimensional systems. Numerical experiments show that E‐DTHPM is better than existing methods and is useful for applications that need to be dynamic enough and accurate enough for fields such as physics, biology, and engineering.

Alzheimer’s disease (AD) is a progressive neurodegenerative disease and the current diagnosis tools, which are used, including clinical examination and neuroimaging, are costly, time consuming and complicated to generalize across demographics. Traditional machine learning (ML) methods such as Support Vector Machines (SVM), LightGBM, and Random Forest (RF) are known to be having weakness in terms of feature selection, class imbalance, and generalization. This paper presents a new Enhanced TabTransformer with Self-Supervised Learning (ETT-SSL) framework of AD classification. The structure combines the transformer-based architecture, self-supervised learning (SSL) for feature representation, SHAP-based feature selection, and the use of SMOTE to balance classes. Experimental results also indicate that ETT-SSL reaches a high accuracy of 95.8%, which is a much better result than baseline models (SVM: 72.1, RF: 78.3, LightGBM: 80.5) and even the standard TabTransformer (85.2%). Additionally, ETT-SSL offers greater accuracy and improved recall, overcoming the issue of false negatives in AD diagnosis. SHAP analysis can be used to improve transparency in clinical decision-making by making use of the model interpretability. The proposed framework recommends a salivary, interpretability and clinically practical methodology that can be used to expand to multimodal data source (e.g. MRI scans, genomic markers and electronic health records) to improve accuracy and generalisability.