Introduction - If you have any usage issues, please Google them yourself
The decoding of Low Density Parity Check (LDPC) codes through the iterative process of belief propagation presents practical challenges for designers looking for real time performance in communication systems. Due to the geometric and finite pattern nature endemic in the construction of LDPC codes, this paper proposes the use of Multi Layer Perception (MLP) feed forward artificial neural networks to replace belief propagation to achieve constant decoding times while retaining performance levels comparable to more traditional decoding methods. Due to the back propagation training method used for neural networks, and the requirement of showing the network every possible input output sequence it will ever see, this paper also presents a novel method of approaching long block length codes far larger than is otherwise possible to train neural networks for with modern computer hardware.