$[16]$

METRICS $=$

BinaryAccuracy $($name$=\text{'}$ 'accuracy'$),$

Precision$($name$=\text{'}$ 'precision'$),$

Recall$($name$=$'recall'$),$

AUC$($name$=\text{'}$ prc$\text{'},$ curve$=\text{'}$PR$\text{'}),$ # precision$-$recall curve

$($a$)(3)$ Next, you will define the deep learning model with BERT and build the model. View the model summary as output.

Your model should include:

$-$an input layer

$-$a BERT preprocessing layer

$-$a BERT layer $($including inputs, BERT encoder, and outputs$)$

$-$a hidden layer with $64$ nodes and reLu activation function

$-$a hidden layer with $28$ nodes and reLu activation function

$-$an output layer with $1$ node and sigmoid activation function

$[]$

$($b$)(3)$ Setup the optimizer as in the BERT paper. Set the initial learning rate to $5e-5.$

$[]$

$($c$)(2)$ Compile and fit your model.

$[]$

$($a$)(2)$ Evaluate the model on both the training and testing sets to evaluate both performance and goodness of fit.