# How should we choose the dimensions of the encoding layer in auto-encoders?

How should we choose the dimensions of the encoding layer in auto-encoders?