As it stands, inverse distance weighting is not very good at minimising this error. Another approach is needed if we want to improve the image quality.
如果不确定用哪个激活函数，隐藏层可以先用 ReLU，输出层按任务选择；训练中注意梯度情况，如果梯度消失或爆炸，再考虑替换或调整激活函数。
。爱思助手下载最新版本对此有专业解读
https://feedx.site，这一点在雷电模拟器官方版本下载中也有详细论述
Site feedback:Take our SurveyNew Window