doi: 10.7763/IJMO.2012.V2.127
Implementation of a Multi-Layer Perceptron Neural Networks in Multi -Width Fixed Point Coding
A.G. Blaiech
,
K. Ben Khalifa
,
M. Boubaker
,
M.H. Bedoui
- TIM Team, Laboratory of Biophysics, Faculty of Medicine of Monastir, University of Monastir, 5019 Tunisia.
Abstract
The artificial neural networks (ANN) have emerged as interesting approaches in various fields of research and industry. ANNs are able to solve non-linear and complex problems where the classic methods do not provide solutions. These learning algorithms are particularly interesting as they are implemented in embedded systems. In order to obtain an efficient implementation, a compromise of time and area is needed In this paper, we will develop a methodology for automatic coding of fixed-point data for the implantation of artificial neural networks that are typically specified as floating-point. This methodology should determine the optimal encoding of various blocks of our ANN, to maximize accuracy and minimize the application area. The proposed methodology allows the automatic generation of VHDL code within an encoding in multi-width of multi-layer perceptron (MLP) model in the decision phase after a simulation in the learning phase with fixed point operators. This methodology has allowed a gain of 6% in LUTs compared to a single coding accuracy with the same network topology.
Keywords
- Floating point
- fixed point
- implementation
- RNA
- simulation
- VHDL
How to Cite
A.G. Blaiech, K. Ben Khalifa, M. Boubaker, and M.H. Bedoui, "Implementation of a Multi-Layer Perceptron Neural Networks in Multi -Width Fixed Point Coding," International Journal of Modeling and Optimization, vol. 2, no. 3, pp. 280-283, 2012. https://doi.org/ 10.7763/IJMO.2012.V2.127
Copyright & License
Copyright © 2012 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).