FIND ME ON

GitHub

LinkedIn

Nearest Neighbour Vector Quantizer (NNVQ)

🌱

Definition
InfoTheory

QQ is a nearest neighbour vector quantizer if for all x\mathbf{x} Q(x)=argmincjC d(x,cj)Q(\mathbf{x})=\underset{ \mathbf{c}_{j}\in\mathcal{C}}{ \arg\min }\ d(\mathbf{x},\mathbf{c}_{j})or equivalently for all i=1,,Ni=1,\dots,N Ri{x:d(x,ci)d(x,cj),j=1,,N}R_{i}\subset \{ \mathbf{x}:d(\mathbf{x},\mathbf{c}_{i})\le d(\mathbf{x},\mathbf{c}_{j}),j=1,\dots,N \}

For the MSE, the overlapping quantizer cells R~i={x:d(x,ci)d(x,cj),j=1,,N}\tilde{R}_{i}=\{ \mathbf{x}:d(\mathbf{x},\mathbf{c}_{i})\le d(\mathbf{x},\mathbf{c}_{j}),j=1,\dots,N \}of a Nearest Neighbour Vector Quantizer (NNVQ) are convex polytopes.

Pasted image 20240402184811.png

Linked from