The cv.jit.learn function made by Jean-Marc Pelletier outputs a statistical distance between the input and the trained model. I know that a statistical distance of 0 would be 100% accurate to the model, but how would you go about turning other values into a probability? or percentage? Say you get a statistical distance of 1; how accurate is that? Could you say that you are 95% confident that it is similar to the trained model?
Or have I completely misunderstood how this works?