After testing the trained neural network, you observe that both training and testing accuracy are quite low. Which of the following new models is MOST LIKELY to provide better accuracy performance?

a$)$ model $=$ Sequential$([$

Dense$(128,$ activation$=$'linear', input $\_$shape$=(64,)),$

Dense$(4,$ activation$=$'linear"$)])$

b$)$ model $=$ Sequential$[$

Dense$(3,$ activation$=$'relu,input$\_$shape $=(3,)),$

Dropout$(0\mathrm{.}2),$

Dense$(2,$ activation$=$'relu'$),$

Dropout $(0\mathrm{.}2),$

Dense$(2,$ activation$=$'relu$),$

Dense$(4,$ activation$=$'linear'$)])$

c$)$ model $=$ Sequential$([$

Dense$(64,$ activation$=$'sigmoid, input shape$=(3),$

Dense$(64,$ activation$=$'sigmoid'$),$

Dense$(64,$ activation$=$'sigmoid"$),$

Dense$(4,$ activation$=$'relu$)])$

d$)$ model $=$ Sequential$($l

Dense$(3,$ activation$=$'relu',input$\_$shape$=(3,)),$

Dropout$(0\mathrm{.}2),$

Dense$(2,$ activation$=$'relu$),$

Dropout$(0\mathrm{.}2),$

Dense$(2,$ activation$=$'relu$),$

Dense$(4,$ activation$=$'sigmoid$)])$