MLMVN Project

2009-2012 - Multilayer Neural Network with Multi-Valued Neurons, its Application to Image Recognition and Processing and Incorporation of the Research Results into the Educational Process.

The project is supported by a $300,000 grant awarded by the National Science Foundation

This "Research in Undergraduate Institutions" project (RUIP) is an integrated research, education, and outreach program that focuses on an in-depth study of complex-valued nonlinear phenomena of a multilayer neural network based on multi-valued neurons (MLMVN). This new original network has a derivative-free self-adaptive learning algorithm and outperforms other artificial neural networks and kernel-based networks in terms of training speed, network complexity and classification/prediction rates. In this RUIP, we will use MLMVN for solving multiple-class classification problems. We concentrate on image recognition problems (texture classification, textural segmentation, blurred image recognition) and intelligent edge detection problems. The goals of the proposed RUIP are:

1) To advance the theory of MLMVN and to comprehensively investigate the complex-valued nonlinear phenomenon of MLMVN. This includes the relationship between the topology of the MLMVN and the quality of multiple-class classification and prediction.

2) To investigate the MLMVN as a multiple-class classifier and the use of the Fourier phase spectrum as a feature space for solving pattern recognition and classification problems.

3) To investigate how the MLMVN can work as a robust edge detector for noisy images and images with a preventing complex background.

4) To develop a hardware implementation of a multi-valued neuron (MVN) and to study a hardware implementation of MLMVN.

5) To implement an educational plan, which incorporates: a) direct involvement of undergraduate students (with special attention to the non-traditional age students) in research through development of undergraduate research projects; b) developing two new undergraduate courses, "Neural Networks and Machine Learning" and "Image Processing and Computer Vision" that will include the novel theories developed within the project along with traditional chapters; c) Public lectures and constructing a web site.