Your colleague has trained a one$-$layer convolution neural network for image classification

using $10000$ sampled images. At the end, he$/$she found that the model performs badly on

both training and test set. What would you suggest, if the aim is to improve predictive

performance?

$($I$)$ Use a more complex pre$-$trained model $($e$.$g$.$ ResNet$-50)$

$($II$)$ Collect more images and retrain the model

$($III$)$ Apply regularisation on your model $($e$.$g$.$ drop$-$out$)$

a$.($I$)$

b$.($III$)$

c$.($I$)$ and $($II$)$

d$.($II$)$ and $($III$)$

e$.($I$),($II$)$ and $($III$)$