Method and apparatus for generating fixed point neural network

A method of updating a fixed point neural network of a neural network system, the fixed point neural network corresponding to a quantized floating point neural network, the method comprising:

• selecting at least one layer of the fixed point neural network as an object layer, wherein the fixed point neural network includes a plurality of layers, each of the plurality of layers corresponding to a respective one of a plurality of quantization parameters;
• forming a candidate parameter set including candidate parameter values with respect to a quantization parameter of the plurality of quantization parameters corresponding to the object layer;
• determining an update parameter value from among the candidate parameter values based on a comparison between levels of network performance of the fixed point neural network, wherein each of the levels of network performance corresponds to a respective one of the candidate parameter values;
• and
• updating the quantization parameter with respect to the object layer based on the update parameter value.