本文构造了两个分别用于分组码的最小软距离译码和最小加权距离译码的前向神经网络,并证明所构造的网络在用于分组码的最小软距离译码和最小加权距离译码时的纠错能力. In this paper, we construct two forward neural networks respectively for minimum soft distance decoding and minimum weighted distance decoding of block codes, and prove that the constructed networks have the best performance in the minimum soft distance decoding and minimum weighted distance decoding Error correction code.