Kevin McFall

Artificial Neural Network Method for Solution of Boundary Value Problems With Exact Satisfaction of Arbitrary Boundary Conditions

A method for solving boundary value problems (BVPs) is introduced using artificial neural networks (ANNs) for irregular domain boundaries with mixed Dirichlet/Neumann boundary conditions (BCs). The approximate ANN solution automatically satisfies BCs at all stages of training, including before training commences. This method is simpler than other ANN methods for solving BVPs due to its unconstrained nature and because automatic satisfaction of Dirichlet BCs provides a good starting approximate solution for significant portions of the domain. Automatic satisfaction of BCs is accomplished by the introduction of an innovative length factor. Several examples of BVP solution are presented for both linear and nonlinear differential equations in two and three dimensions. Error norms in the approximate solution on the order of 10-4 to 10-5 are reported for all example problems.

Automated Design Parameter Selection for Neural Networks Solving Coupled Partial Differential Equations with Discontinuities

Abstract The length factor artificial neural network (ANN) method for solving coupled systems of partial differential equations (DEs) is unique among ANN methods in that the approximate solution exactly satisfies boundary conditions (BCs) on arbitrary geometries regardless of the ANN output. Besides removing the BC constraint from the optimization process, this property allows the method to accurately solve problems with discontinuous BCs despite the continuous nature of ANNs. An automated design parameter selection process is developed to choose a single ANN from an ensemble comprising numerous combinations of design parameters and random starting weights and biases. The selection is made completely independently of the human designer by comparing the magnitude and uniformity of each approximate solutions error in satisfying the DE(s). The automated selection process successfully chooses a solution with error on the same order of magnitude as the best solution in the ensemble. The resulting approximations provide low error solutions for the three different thermal-fluid science example problems explored, including the Navier–Stokes equations.

Software Effort Estimation Using a Neural Network Ensemble

Accurate software effort estimation is crucial for software consulting organizations to stay competitive in their software development costs and retain customers. Artificial Neural Network (ANN) is an effective tool to obtain accurate effort estimates. In this paper, software effort estimation models using Artificial Neural Network (ANN) ensembles and regression analysis are developed based on data collected from 163 software development projects. The main emphasis of the paper is in developing an effective experimental design to achieve superior effort estimation results. In addition, we compare the software effort estimation of ANNs and multiple regression analysis. We found two interesting results. First, variables other than size (function points) are not especially helpful in predicting software development effort. Second, a properly designed ANN ensemble significantly outperforms estimation using regression analysis and can achieve better effort estimate predictions.

A length factor artificial neural network method for the numerical solution of the advection dispersion equation characterizing the mass balance of fluid flow in a chemical reactor

In this article, a length factor artificial neural network (ANN) method is proposed for the numerical solution of the advection dispersion equation (ADE) in steady state that is used extensively in fluid dynamics and in the mass balance of a chemical reactor. An approximate trial solution of the ADE is constructed in terms of ANN using the concept of the length factor in a way that automatically satisfies the desired boundary conditions, regardless of the ANN output. The mathematical model of ADE is presented adopting a first-order reaction, and the steady-state case for the same is examined by estimating the numerical solution using the ANN technique. Numerical simulations are performed by choosing the best ANN ensemble, based on a combination of numerous design parameters, random starting weights, and biases. The solution obtained using the ANN method is compared to the existing finite difference method (FDM) to test the reliability and effectiveness of the proposed approach. Three cases of ADE are considered in this study for different values of advection and dispersion. The numerical results show that the ANN method exhibits a higher accuracy than the FDM, even for the smaller number of training points in the domain, and eliminates the instability issues for the case where advection dominates dispersion.

Using Visual Lane Detection to Control Steering in a Self-Driving Vehicle

An effective lane detection algorithm employing the Hough transform and inverse perspective mapping to estimate distances in real space is utilized to send steering control commands to a self-driving vehicle. The vehicle is capable of autonomously traversing long stretches of straight road in a wide variety of conditions with the same set of algorithm design parameters. Better performance is hampered by slowly updating inputs to the steering control system. The 5 frames per second (FPS) using a Raspberry Pi 2 for image capture and processing can be improved to 23 FPS with an Odroid XU3. Even at 5 FPS, the vehicle is capable of navigating structured and unstructured roads at slow speed.

Stimulating Class Discussion Using an Online Newspaper Created with Twitter and Paper.li.

Courses: First-year Experience, Intercultural Community Building, and Introductory Communication Objective: Students develop a sense of community through this assignment by discussing personally relevant topics via the shared experience enabled by the social media tools Twitter and Paper.li.

Future heat transfer concerns in Josephson junction computers

Using the superconducting properties of Josephson junctions enable extremely high switching speeds unmatched in semiconducting electronics. Much research has been conducted in recent decades in order to produce high performance electronics based on Josephson junction logic. In addition to the high speeds attainable by this technology, also of significance is the very low heat dissipated by Josephson circuits. Josephson devices have made great strides in the last ten years with microprocessors reaching levels of integration as high as 10/sup 5/ junctions/cm/sup 2/. Dissipation in these devices is easily managed, but integrations reaching 10/sup 7/ must be considered if Josephson electronics are to compete with the complexity and functionality of semiconducting electronics. Coupling this level of integration with dissipations of 0.34 and 2.98 /spl mu/W/junction in low and high temperature cases respectively, produces large heat fluxes difficult to remove at cryogenic temperatures. While other technical difficulties currently overshadow heat transfer concerns, the future of Josephson electronics research will likely need to address them.