Hello guys,
I am Menelaos from Greece.
After plenty of studying, notes’ preparations and mental clarity I re-enrolled on Sunday in Course 2 of Advanced Learning Algorithms.
Today I managed to reach Week 3 of the graded Labs having reached almost in the middle.
My issue is the highly intuitive formulas of the graded cell eval_cat_err.
I have attached screenshots of how I thought of it but I’m still confused. We have to calculate a branched sum trying to figure out how many times predictions are identical and how many times they differ. The sum of the constant 1 from 0 to (m-1) equals 1*m. So we could say that they differ (m-TruePositives) times and they’re identical (TruePositives) times. But that’s just theory cos calculations and variables differ, it’s just theory help.
If they differ (yhat - y) is not 0. If they’re identical, as indices, (yhat-y) is 0. Or we can use the fraction of (yhat/y) to decide, meaning, being 1 or not 1. We’re talking multiclass classification and we have to discover a mechanical analogy of the computations that gives us the exact proportions. I’m dazzled and can’t figure out the equations.
Can someone, instructor or mentor who knows the problem guide me to how am I supposed to pass it? I would truly appreciate an not on-the-go answer.
Looking forward to hearing from you
Best
M
{mentor edit: code image removed}