摘要 |
PURPOSE: A neural network decoder is provided to simplify a structure of a decoder and increase accuracy in a decoding process by forming the decoder using a neural network. CONSTITUTION: The device includes a switch(602), toggle latch(604), input buffer(601), output process part(620), bit register(622), and 4-port-2 throw switch(624). A method for training a neural network includes a process for providing an input symbol, a comparing process, and a difference reduction process. The process for providing the input symbol is performed to provide a plurality of continuous input symbols to a neural network and an encoder. The comparing process is performed to compare an output signal of the neural network with the input signal. A difference reduction process is performed to select a neural network variable in order to reduce a difference between neural network variables.
|