parametric model is designed to solve a supervised learning $-$ binary classification

problem. The problem has n input features and a binary output y in $\{0,1\}.$ The model

would compute a predicted probability function p : Rn $(0,1)$ with p$($v$)$ representing

the probability that the input v belongs to class y $=1.$

Consider the following data:.

$1$

v y pA pB

v$110\mathrm{.}950\mathrm{.}90$

v$200\mathrm{.}350\mathrm{.}40$

v$310\mathrm{.}700\mathrm{.}80$

v$410\mathrm{.}600\mathrm{.}55$

v$510\mathrm{.}800\mathrm{.}77$

The first two columns from the left provide a training dataset of size m $=5.$

One considers two distinct parameter sets, to be possibly used by the model:

$$ Parameter set A leads to a probability function pA : Rn $(0,1).$ The values of pA

on the training inputs are given in column three $($labeled pA$).$

$$ Parameter set B leads to a probability function pB : Rn $(0,1).$ The values of pB

on the training inputs are given in column three $($labeled pB $).$

The model uses Binary Cross$-$Entropy as loss function. Between the two parameters A

and B$,$ which one is preferable for the model to use?