The current progress in **image recognition** is made by changing the approach from a classic feature selection (shallow learning algorithm) to no feature selection (deep learning), due to the mathematical properties of** convolutional neural networks**. Their ability to capture the same information using a smaller number of parameters was partially caused by their shift-invariance property.

The main reason behind this success was developing faster learning algorithms than more mathematically accurate ones, so less attention is put on developing another property invariant neural nets.

Hope this answer helps.