The activation function used to train the network is given as f$($x$)=1/(1+$e$^(-($w$.$x$+$b$)))$ and the loss function used is squared error loss function. With respect to this, using gradient descent algorithm show how the values of $$w$$ and $$b$$ varies if the initial setup is:

i$)$w$$ is chosen as $-$ve$$ and $$b$$ is $-$ve$.$

ii$)$w$$ is chosen $+$ve and $$b$$ is also chosen $+$ve$.$

iii$)$w$$ is chosen $+$ve and $$b$$ is chosen $-$ve$.$