This work proposes deep learning based approach for decoding of Turbo product codes(TPCs). Deep Neural Network (DNN) decoder is used as the Soft-Input Soft-Output (SISO)decoder in the iterative decoding of product codes. The method implements one shot decoding thus enabling high level of parallelism. Due to the highly parallelizable nature of the DNN based SISO decoder, the computational complexity and time complexity are lowered compared to the conventional Chase SISO decoder. This in turn reduces the overall decoding complexity for TPCs. The DNN decoder, which is based on belief propagation algorithm trains the weights assigned over edges of Tanner graph. Simulation results show that the proposed decoder can achieve performance similar to that of conventional Chase-Pyndiah algorithm. The proposed method finds use in data storage and multimedia applications which has stringent requirements for high data rate, low decoding delay and low decoding complexity.
