Advanced optimization to unroll parameter from matrices into vectors
theta becomes vector, not a single unit
unroll so advanced optimization can receive vectors of theta as an input
Reshape: takes number of elemen from a vector, 1:110 (Because theta1 = 10x11) and reshape as one matrices with 10x11 shape
so that's it for matrix vs vector representation
Advantage matrices: more convenient forward back prop for vectorized implementation
Advantage vector = vectorized assume dvec,thetavec as just one big vector