Dun Pu’s research while affiliated with Tibet University and other places

Both all-connection model and part connection model is simulated, which adopt self-organizing map neural network to generate check bits. Consequently, N source bits and K check bits are composed a complete codeword. In the decoding port, multi-layer perceptron network (MLPN) is utilized to implement decoding function. The specific steps are as follows: (1) Constructing the MLPN according to the size of codeword sets and source bits; (2) Training the MLPN with codeword sets generated by neural network decoder until qualified; (3) Accepting and decoding codeword sets via trained MLPN. Actual tests show that: (1) There exist no evident performance differences between all-connection model and part-connection model; (2) The connection of weight sets is similar to Tanner graph in part-connection model, which reduce the computational complex greatly and remain good performance at the same time. In sum, the method of encoding and decoding has certain market prospect in confidential communication field.

The paper presents a novel method of initial weights optimization method in Multi-Layer Perceptron Network(MLPN). Firstly, the sample sets should be transformed by K-L Transform. Secondly, use K-L Converting Matrix to initialize the weights between input and hidden layer. Thirdly the MLPN is trained by BP algorithm, and the convergence speed of MLPN is improved evidently. The ultimate test shows the new algorithm is suitable for the situation of low-dimensional data.