Q$.2$ Perceptron $/$ NN $/$ Q$-$learning $($check the attached screenshot$)($a$)$ Explain the difference between online learning and batch learning.

$($b$)$ Consider the following perceptron with $3$ inputs, one output, and a step function as the activation

function, with a threshold value of $0.$ The bias is initially $\{b=0\mathrm{.}5\}.$ Given the inputs $\{i0=1,i1=0,i2=0\}$

the output is $\{00=1\}.$ The activation is equal to $i{0}^{**}w0+i{1}^{**}w1+i{2}^{**}w2+b.$ Derive the value$($s$)$ of the

weight w$0?$

$($c$)$ Describe the main steps of the supervised training algorithm for a multi$-$layer feed$-$forward neural

network.

$($d$)$ Sketch the sigmoid function. Is it continuous? Why is it useful for training with gradient descent?

$($e$)$ What is the difference between on$-$policy and off$-$policy methods?

$($f$)$ Explain briefly the Q$-$learning algorithm