When we started to do something about machine learning, we read many papers, written by various background, who used totally different notations. It is a rather hard time to understand a same thing when my brain needs to frequently shift from a set of notation to another one. In addition, it is extremely hard to choose notations when we writing the paper even for simple concepts, such as sample number and neuron number.
When I listen to talks, it is really a burden to remember the notations.
Based on such experience, we do feel a suggestion of standard notation is necessary. This suggestion may be not complete and may have some conflicts at current time, but it is a good sign if the community are working hard towards a standard of notation usage.