Penggang GAO

Thank you so much. It works.

微调时，DBN的每一层训练完成后自然要把参数传递给一个大的NN，这个NN的输入时什么？结构是否会应为要继承RBM的w.b 受限？损失函数如何定义？我是研究机器人相关的，是dl的初学者，请见谅。

哈哈，应该是我问的有问题。我自己查资料吧，谢谢了。

I also tried loam-livox on Mid-70 and it can not work well. But livox_mapping performed well based on Mid-70. Is this problem caused by Mid-70？

download turtlebot_gazebo,then catkin_make again