Not that I like it, but a lot of algorithms in papers use shorthand subscripting so they can write equations that aren't 300 columns long. Say you're doing neur

Not that I like it, but a lot of algorithms in papers use shorthand subscripting so they can write equations that aren't 300 columns long. Say you're doing neural network stuff like w_1,w_2,w_3,...,w_n for weights, inputs, outputs, etc. If you're reproducing those algorithms in a program, you could do worse than copying their naming structure so that the correspondence between the program and the paper is as obvious as possible.