Supervised Learning with Potential Functions for Neural Network-Based Object Recognition

Project Goal

The goal of this project is to develop supervised learning algorithms for feed forward and RBF neural networks and a novel method for data clustering which performs classification based on a set of potential fields synthesized over the domain on input space by a number of potential function units.

Expected Outcomes

• To increase the confidence of women students and to help create a positive and empowering environment for women where they can continue to grow both professionally and personally.

• The students will extend their mathematical background; enhance their computing skills and expertise by introducing them to state-of-the-art hardware and software in object-oriented technology, simulation and modeling.

• The participating students will be introduced to team-based and cross-disciplinary research.

• To help students prepare for graduate school and define career goals and to encourage them to pursue careers as research scientists.

• Two neural network architectures (feed forward and radial basis functions) based on a new approach -potential functions will be designed.

• New supervised learning algorithms allowing structural changes in hidden layers for both neural network architectures will be developed and implemented.

• The neural network topologies will include a set of different orthogonal functions allowing the user to make a choice and to analyze the neural network output for different training sets.

• The results will be submitted for possible presentation and publication at international conferences.