While it is straightforward to invert a function like y = mx to produce the inverse, x = y/m, some functions can’t be easily inverted. One such function is the cumulative distribution function (CDF) of the normal probability distribution, where neither the CDF nor the inverse CDF (quantile function) can be expressed in a closed form. This article presents a method for inverting a function using a neural network regardless of whether the problem can be solved analytically.
