Training of CSRN
Description/Abstract/Artist Statement
In the past decade researchers put significant effort to solve complex problems using Artificial Neural Networks (ANNs). A unique type of ANN known as a Cellular Simultaneous Recurrent Network (CSRN) has shown excellent performance for solving complex topological mapping problems such as maze traversal. However, the complexity of CSRN architecture makes it very hard to train. Researchers in ODU’s Vision Lab have improved the training by introducing Unscented Kalman Filter (UKF) algorithm but it is not yet real time. In this research we plan to implement CSRN in a hardware platform to achieve real time performance for large scale problems.
Presentation Type
Event
Disciplines
Electrical and Computer Engineering
Location
Learning Commons @ Perry Library, Northwest Atrium
Start Date
2-13-2016 8:00 AM
End Date
2-13-2016 12:30 PM
Upload File
wf_no
Training of CSRN
Learning Commons @ Perry Library, Northwest Atrium
In the past decade researchers put significant effort to solve complex problems using Artificial Neural Networks (ANNs). A unique type of ANN known as a Cellular Simultaneous Recurrent Network (CSRN) has shown excellent performance for solving complex topological mapping problems such as maze traversal. However, the complexity of CSRN architecture makes it very hard to train. Researchers in ODU’s Vision Lab have improved the training by introducing Unscented Kalman Filter (UKF) algorithm but it is not yet real time. In this research we plan to implement CSRN in a hardware platform to achieve real time performance for large scale problems.
Comments
Faculty Mentor: Dr. Khan M. Iftekharuddin