深度学习中的CNN方法，层数由原先的五层增加到七层，使用最新的RELU激活函数代替原先的sigm函数，能运行-CNN depth learning method, the number of layers the original five to seven, with the latest RELU activation function replaces sigm function, you can run it

实现CNN的手写字体matlab程序，可调节激活函数，选择loss函数(Realization of handwritten font matlab CNN procedures, can regulate the activation function, loss function)