2 ms·
A confidence interval is a sort of meaningless notion in this case. You can just do the derivative of the algorithm's answer, and see if you get back an express
by bglazer 6y ago
A confidence interval is a sort of meaningless notion in this case. You can just do the derivative of the algorithm's answer, and see if you get back an expression that is equivalent to the input. If you did, then the answer is correct.
- clircle 6y agoI don't believe a model (neural network or otherwise) is capable of knowing that its predictions are correct.