How do autoencoders came to know that which features are most salient?

Question

Autoencoders are generally used for reducing the number of dimensions. It compresses the image by reducing the unnecessary number of dimensions . . They work by compressing the