An interesting equality from linear algebra
It’s a bit mind-blowing when you want to understand this equality geometrically:
we have vectors $x_1, …, x_n$ in n-dimensional space
Let’s take an orthonormal basis in this space $e_1, …, e_n$ and compute such vectors obtained with via scalar product:
So the theorem is:
\[\det_{i,j} < z_i, z_j > = \det_{i,j} < x_i, x_j >\]And the proof is very simple, let’s introduce the matrix $A$: $ A_{ij} = <x_i, e_j> $
\[\det_{i,j} < z_i, z_j > = \det A^T A = \det A A^T = \det_{i,j} < x_i, x_j >\]Voila! Beautiful, but completely unclear.
Gradient boosting
Hamiltonian MC
Gradient boosting
Reconstructing pictures
Neural Networks