Bit misleading, parameters here are already obtained - the only lisp part is applying pre-obtained weights to an input and computing the output (i.e. no training here).

That aside, this is very cool nevertheless.