Nearest-Neighbour-Search-Neural-Networks-Product-Quantization/MNIST/Product Quantization/RunPQNNSwithADC.m

clear all
close all 
clc

kk=2.^[1:12];
rat=[1 2 5 10 20 50 100];
mm=[4 7 16];
%kk=2048;
%mm=[8 16];
nRecall=100;

%%

for ii=1:length(mm)
    for jj=1:length(kk)
        [Performances(jj,ii,:) normCCostADC(jj,ii), ~] = PQNNSwithADC( kk(jj), mm(ii), nRecall, [0 0 0 0 1])
    end
end

%%
for jj=1:length(mm)
figure
for rr=1:length(rat)
    plot(normCCostADC(:,jj),Performances(:,jj,rr),'-x'), hold on
end
for ii=1:length(kk)
    text(normCCostADC(ii,jj)+2e-5,Performances(ii,jj,1)-.01,sprintf('$k=%d$',kk(ii)),'interpreter','latex');
end
h=legend('recall@1','recall@2','recall@5','recall@10','recall@20','recall@50','recall@100');
set(h,'Interpreter','latex');
grid on; grid minor;
title(sprintf('Product Quantization with ADC - $m=%d$',mm(jj)),'interpreter','latex');
xlabel('Computational Cost (scaled to Exhaustive one)','interpreter','latex');
ylabel('$\eta_r$','interpreter','latex');
end

%%