Characterized by their high spatiotemporal variability, rainfalls are difficult to predict, especially under climate change. This study proposes a multilayer perceptron (MLP) network optimized by Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Firefly Algorithm (FFA), and Teleconnection Pattern Indices - such as North Atlantic Oscillation (NAO), Southern Oscillations (SOI), Western Mediterranean Oscillation (WeMO), and Mediterranean Oscillation (MO) - to model monthly rainfalls at the Sebaou River basin (Northern Algeria). Afterward, we compared the best-optimized MLP to the application of the Extreme Learning Machine optimized by the Bat algorithm (Bat-ELM). Assessment of the various input combinations revealed that the NAO index was the most influential parameter in improving the modeling accuracy. The results indicated that the MLP-FFA model was superior to MLP-GA and MLP-PSO for the testing phase, presenting RMSE values equal to 33.36, 30.50, and 29.92 mm, respectively. The comparison between the best MLP model and Bat-ELM revealed the high performance of Bat-ELM for rainfall modeling at the Sebaou River basin, with RMSE reducing from 29.92 to 11.89 mm and NSE value from 0.902 to 0.985 during the testing phase. This study shows that incorporating the North Atlantic Oscillation (NAO) as a predictor improved the accuracy of artificial intelligence systems optimized by metaheuristic algorithms, specifically Bat-ELM, for rainfall modeling tasks such as filling in missing data of rainfall time series.

The use of models capable of forecasting the production of photovoltaic (PV) energy is essential to guarantee the best possible integration of this energy source into traditional distribution grids. Long Short-Term Memory networks (LSTMs) are commonly used for this purpose, but their use may not be the better option due to their great computational complexity and slower inference and training time. Thus, in this work, we seek to evaluate the use of neural networks MLPs (Multilayer Perceptron), Recurrent Neural Networks (RNNs), and LSTMs, for the forecast of 5 min of photovoltaic energy production. Each iteration of the predictions uses the last 120 min of data collected from the PV system (power, irradiation, and PV cell temperature), measured from 2019 to mid-2022 in Maceió (Brazil). In addition, Bayesian hyperparameters optimization was used to obtain the best of each model and compare them on an equal footing. Results showed that the MLP performs satisfactorily, requiring much less time to train and forecast, indicating that they can be a better option when dealing with a very short-term forecast in specific contexts, for example, in systems with little computational resources.

Revealing associations among various structural and functional patterns of the brain can yield highly informative results about the healthy and disordered brain. Studies using neuroimaging data have more recently begun to utilize the information within as well as across various functional and anatomical domains (i.e., groups of brain networks). However, most whole-brain approaches assume similar complexity of interactions throughout the brain. Here we investigate the hypothesis that interactions between brain networks capture varying amounts of complexity, and that we can better capture this information by varying the complexity of the model subspace structure based on available training data. To do this, we employ a Bayesian optimization-based framework known as the Tree Parzen Estimator (TPE) to identify, exploit and analyze patterns of variation in the information encoded by temporal information extracted from functional magnetic resonance imaging (fMRI) subdomains of the brain. Using a repeated cross-validation procedure on a schizophrenia classification task, we demonstrate evidence that interactions between specific functional subdomains are better characterized by more sophisticated model architectures compared to less complicated ones required by the others for optimally contributing towards classification and understanding the brain\'s functional interactions. We show that functional subdomains known to be involved in schizophrenia require more complex architectures to optimally unravel discriminatory information about the disorder. Our study points to the need for adaptive, hierarchical learning frameworks that cater differently to the features from different subdomains, not only for a better prediction but also for enabling the identification of features predicting the outcome of interest.

A critical step in the production of Esomeprazole powder for solution is a period between the filling process and lyophilization, where all vials, partially closed, are completely exposed to environmental influences. Excessive instability reflects in pH value variations caused by oxygen\'s impact. In order to provide pH control, which consequently affects drug stability, Esomeprazole batches, produced in the same way, were kept in partially closed vials for 3 h at temperatures of 20 °C and -30 °C, after which they were lyophilized and stored for long-term stability for 36 months. The aim of the presented study was to apply a deep-learning algorithm for the prediction of the Esomeprazole stability profile and to determine the pH limit for the reconstituted solution of the final freeze-dried product that would assure a quality product profile over a storage period of 36 months. Multilayer perceptron (MLP) as a deep learning tool, with four layers, was used. The pH value of Esomeprazole solution and time of storage (months) were inputs for the network, while Esomeprazole assay and four main impurities were outputs of the network. In order to keep all related substances and Esomeprazole assay in accordance with specifications for the whole shelf life, the pH value for the reconstituted finish product should be set in the range of 10.4-10.6.

BACKGROUND: Optimizing the somatic embryogenesis protocol can be considered as the first and foremost step in successful gene transformation studies. However, it is usually difficult to achieve an optimized embryogenesis protocol due to the cost and time-consuming as well as the complexity of this process. Therefore, it is necessary to use a novel computational approach, such as machine learning algorithms for this aim. In the present study, two machine learning algorithms, including Multilayer Perceptron (MLP) as an artificial neural network (ANN) and support vector regression (SVR), were employed to model somatic embryogenesis of chrysanthemum, as a case study, and compare their prediction accuracy.
RESULTS: The results showed that SVR (R2 > 0.92) had better performance accuracy than MLP (R2 > 0.82). Moreover, the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) was also applied for the optimization of the somatic embryogenesis and the results showed that the highest embryogenesis rate (99.09%) and the maximum number of somatic embryos per explant (56.24) can be obtained from a medium containing 9.10 μM 2,4-dichlorophenoxyacetic acid (2,4-D), 4.70 μM kinetin (KIN), and 18.73 μM sodium nitroprusside (SNP). According to our results, SVR-NSGA-II was able to optimize the chrysanthemum\'s somatic embryogenesis accurately.
CONCLUSIONS: SVR-NSGA-II can be employed as a reliable and applicable computational methodology in future plant tissue culture studies.

Over the past few decades, geometric morphometric methods have become increasingly popular and powerful tools to describe morphological data while over the same period artificial neural networks have had a similar rise in the classification of specimens to preconceived groups. However, there has been little research into how well these two systems operate together, particularly in comparison to preexisting techniques. In this study, geometric morphometric data and multilayer perceptrons, a style of artificial neural network, were used to classify shark teeth from the genus Carcharhinus to species. Three datasets of varying size and species differences were used. We compared the performance of this combination with geometric morphometric data in a linear discriminate function analysis, linear measurements in a linear discriminate function analysis, and a preexisting methodology from the literature that incorporates linear measurements and a two-layered discriminate function analysis. Across datasets, geometric morphometric data in a multilayer perceptron tended to yield modest accuracies but accuracies that varied less across species whereas other methods were able to achieve higher accuracies in some species at the expense of lower accuracies in others. Further, the performance of the two-layered discriminate function analysis illustrates that constraining what material is classified can increase the accuracy of a method. Based on this tradeoff, the best methodology will then depend on the scope of the study and the amount of material available. J. Morphol. 278:131-141, 2017. ©© 2016 Wiley Periodicals,Inc.

Prioritization of compounds using inverse docking approach is limited owing to potential drawbacks in its scoring functions. Classically, molecules ranked by best or lowest binding energies and clustering methods have been considered as probable hits. Mining probable hits from an inverse docking approach is very complicated given the closely related protein targets and the chemically similar ligand data set. To overcome this problem, we present here a computational approach using receptor-centric and ligand-centric methods to infer the reliability of the inverse docking approach and to recognize probable hits. This knowledge-driven approach takes advantage of experimentally identified inhibitors against a particular protein target of interest to delineate shape and molecular field properties and use a multilayer perceptron model to predict the biological activity of the test molecules. The approach was validated using flavone derivatives possessing inhibitory activities against principal antimalarial molecular targets of fatty acid biosynthetic pathway, FabG, FabI and FabZ, respectively. We propose that probable hits can be retrieved by comparing the rank list of docking, quantitative-structure activity relationship and multilayer perceptron models.