SRNEngine: A Windows-based Neural Network Simulation Tool for the Non-programmer

Abstract

SRNEngine is a windows-based application package for training neural networks. The graphical user interface allows the drag-and-drop creation of neural networks with a variety of architectures, without the need for any programming. At present, these architectures/learning algorithms include Simple Recurrent Networks, Jordan networks, and any kind of feedforward backpropagation network, with up to five each of input, hidden, and output layers (pools of units). A version that adds backpropagation-through-time is in development. The interface is designed to conform to the Microsoft Windows GUI environment that most PC users are already familiar with. SRNEngine includes tools for creating, editing, and manipulating various types of training data, and is especially optimized for working with text/language data, including automatic word-to-input-representation translation at runtime for text corpora. The distributed computing feature allows multiple simulations to be run on a network of workstations, co-ordinated via a central ftp server.


Back to Table of Contents