Liver cancer, one of the rapidly escalating forms of cancer, remains a principal cause of mortality globally. Its death rates can be attenuated through vigilant monitoring and early detection. This study aims to develop a sophisticated model to assist medical professionals in the classification of liver tumours using biopsy tissue images, thereby facilitating preliminary diagnosis.The study presents a novel, bio-inspired deep learning strategy purposed for augmenting liver cancer detection. The uniqueness of this approach rests in its two-fold contribution: Firstly, an innovative hybrid segmentation technique, integrating the SegNet network, UNet network, and Al-Biruni Earth Radius (BER) procedure, is introduced to extract liver lesions from Computed Tomography (CT) images. The algorithm initially applies the SegNet to isolate the liver from the abdominal image in a CT scan. Since hyperparameters significantly influence segmentation performance, the BER algorithm is hybridized with each network for optimal tuning. The method proposed herein is inspired by the pursuit of a common objective by swarm members. Al-Biruni's methodology for calculating Earth's radius sets the search space, extending beyond local solutions that require exploration. Secondly, a pre-trained AlexNet model is utilized for diagnosis, further enhancing the method's effectiveness. The proposed segmentation and classification algorithms have been compared with contemporary state-of-the-art techniques. The results demonstrated that in terms of specificity, F1-score, accuracy, and computational time, the proposed method outperforms its competitors, indicating its potential in advancing liver cancer detection.

Pharmaceutical transport logistics, especially in humanitarian and hospital contexts, is becoming increasingly essential with a growing need to monitor associated costs. In Morocco, however, studies focusing on the cost implications of pharmaceutical delivery conditions are conspicuously absent. This creates a high-dimensional classification framework, where the selection of variables becomes challenging in the face of correlated distribution predictors. The integration of Artificial Intelligence (AI) in cost prediction has emerged as a vital necessity amidst escalating complexities and cost considerations. Cost prediction, being inherently correlated with almost all variables and inputs, offers an interpretable value in performance management, financial planning, and contract negotiation. This study undertakes a comparative analysis of a broad spectrum of prediction algorithms applied to the same, albeit reduced, database. A dozen such algorithms are put into practical use, with variable selection implemented through importance measures. The primary objective of this comparative evaluation is to determine the superior performing algorithm — one that delivers optimal adaptation to the context within a fixed environment. The prediction algorithm incorporates a myriad of inputs and constraints derived from data collection systems. AI's application facilitates the inclusion of diverse variables such as transportation routes, congestion, distances, freight weight, and environmental factors, thereby enhancing the accuracy and efficiency of cost estimation. The Orthogonal Matching Pursuit model emerged as the most successful, boasting an R² value nearing unity. Accurate cost prediction in transport can yield valuable insights into budgeting, estimation, customer service, managerial risk, environmental considerations, and strategic deployment for a company. Improved decision-making and resource allocation can thereby be achieved, leading to enhanced profitability and sustainability.

Dominant points, or control points, represent areas of high curvature on shape contours and are extensively utilized in the representation of shape outlines. Herein, we introduce a novel, descriptor-based approach for the efficient detection of these pivotal points. Each point on a shape contour is evaluated and mapped to an invariant descriptor set, accomplished through the use of point-neighborhood. These descriptors are then harnessed to discern whether a point qualifies as a dominant one. Our proposed methodology eliminates the need for costly computations typically associated with evaluating candidate dominant points. Furthermore, our algorithm significantly outperforms its predecessors in terms of speed, relying solely on integer operations and obviating the necessity for an optimization phase. Experimental outcomes, derived from the widely used MPEG7_CE-Shape-1_Part_B, denote a minimum enhancement of 2.3 times in terms of running time. This implies that the proposed methodology is particularly suitable for real-time applications or scenarios managing shapes comprising a substantial number of points.

The task of interpreting multi-variable time series data, while also forecasting outcomes accurately, is an ongoing challenge within the machine learning domain. This study presents an advanced method of utilizing Long Short-Term Memory (LSTM) recurrent neural networks in the analysis of such data, with specific attention to both target and exogenous variables. The novel approach aims to extract hidden states that are unique to individual variables, thereby capturing the distinctive dynamics inherent in multi-variable time series and allowing the elucidation of each variable's contribution to predictive outcomes. A pioneering mixture attention mechanism is introduced, which, by leveraging the aforementioned variable-specific hidden states, characterizes the generative process of the target variable. The study further enhances this methodology by formulating associated training techniques that permit concurrent learning of network parameters, variable interactions, and temporal significance with respect to the target prediction. The effectiveness of this approach is empirically validated through rigorous experimentation on three real-world datasets, including the 2022 closing prices of three major stocks - Apple (AAPL), Amazon (AMZN), and Microsoft (MSFT). The results demonstrated superior predictive performance, attributable to the successful encapsulation of the diverse dynamics of different variables. Furthermore, the study provides a comprehensive evaluation of the interpretability outcomes, both qualitatively and quantitatively. The presented framework thus holds substantial promise as a comprehensive solution that not only enhances prediction accuracy but also aids in the extraction of valuable insights from complex multi-variable datasets.

In addressing the challenge of obstacle scattering inversion amidst intricate noise conditions, a model predicated on convolutional neural networks (CNN) has been proposed, demonstrating high precision. Five distinct noise scenarios, encompassing Gaussian white noise, uniform distribution noise, Poisson distribution noise, Laplace noise, and impulse noise, were evaluated. Far-field data paired with the Fourier coefficients of obstacle boundary curves were employed as network input and output, respectively. Through the convolutional processes inherent to the CNN, salient features within the far-field data related to obstacles were adeptly identified. Concurrently, the statistical characteristics of the noise were assimilated, and its perturbing effects were diminished, thus facilitating the inversion of obstacle shape parameters. The intrinsic capacity of CNNs to intuitively learn and differentiate salient features from data eradicates the necessity for external intervention or manually designed feature extractors. This adaptability confers upon CNNs a significant edge in tackling obstacle scattering inversion challenges, particularly in light of fluctuating data distributions and feature variability. Numerical experiments have substantiated that the aforementioned CNN model excels in addressing scattering inversion complications within multifaceted noise conditions, consistently delivering solutions with remarkable precision.