Filters

20222
Reset filters

Settings

Subject: Artificial neural networks

Search Results

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Artificial neural network-based heuristic to solve COVID-19 model including government strategies and individual responses
    (Informatics in Medicine Unlocked, 2022-07-16) Thongchai Botmart; Zulqurnain Sabir b; Shumaila Javeed c; Rafaél Artidoro Sandoval Núñez d; Wajaree weera a; Mohamed R. Ali e; R. Sadat f
    The current work aims to design a computational framework based on artificial neural networks (ANNs) and the optimization procedures of global and local search approach to solve the nonlinear dynamics of the spread of COVID-19, i.e., the SEIR-NDC model. The combination of the Genetic algorithm (GA) and active-set approach (ASA), i.e., GA-ASA, works as a global-local search scheme to solve the SEIR-NDC model. An error-based fitness function is optimized through the hybrid combination of the GA-ASA by using the differential SEIR-NDC model and its initial conditions. The numerical performances of the SEIR-NDC nonlinear model are presented through the procedures of ANNs along with GA-ASA by taking ten neurons. The orrectness of the designed scheme is observed by comparing the obtained results based on the SEIR-NDC model and the reference Adams method. The absolute error performances are performed in suitable ranges for each dynamic of the SEIR-NDC model. The statistical analysis is provided to authenticate the reliability of the proposed scheme. Moreover, performance indices graphs and convergence measures are provided to authenticate the exactness and constancy of the proposed stochastic scheme.
  • Thumbnail Image
    Item
    IoT technology enabled stochastic computing paradigm for numerical simulation of heterogeneous mosquito model
    (springer Link, 2022-11-19) Sohaib Latif; Zulqurnain Sabir; Muhammad Asif Zahoor Raja; Cieza Altamirano, Gilder; Sandoval Núñez, Rafaél Artidoro; Oseda Gago, Dulio; R. Sadat; Mohamed R. Ali
    In this communication, a fractional order design and numerical form of the solutions are presented for numerical simulations of heterogeneous mosquito model. The use of the fractional order derivatives is exploited to observe more accurate and exhaustive performances of the numerical simulation of the model. The novel design of the fractional order heterogeneous mosquito differential system is analyzed with stochastic solver based on the internet of things (IoT) technologies, represented with four categories i.e., normal individuals, people with reflex behavior, panic behavior and controlled behavior based differential system. The solutions of the novel design of the fractional order system are presented by using the stochastic paradigm of artificial neural network (ANN) procedures along with the Scaled Conjugate Gradient (SCG), i.e., ANN-SCG, for learning of weights. In ANN-SCG implementation, the data statistics are picked as 78% for training, 11% for both authorization and testing samples to approximate the solutions. The accuracy of the ANN-SCG technique is seen by correlation of the determined outcomes and the information base on Adams-Bashforth-Moulton method based standard solutions. To achieve the capacity, legitimacy, consistent quality, fitness, and accuracy of the ANN-SCG strategy, the reproductions-based error histograms (EHs), MSE, regression, and state transitions (STs) are used for extensive experimentations.