Q$3.$ Train the perceptron network given the implemented learning rule. First, define the following as your training set.

Then, initialize the weights for your network randomly.

Next, pass the first data point to your perceptron, to get the predicted output $($a$).$

Next, pass the following to the learning rule model: a $($predicted output$),t($actual target output$),w($current weight$)$ and b $($current bias$).$

The learning rule function will return the update weight and bias.

Repeat this step for each data point in your training set.

Stop when $w$ and b converges.

You may need to iterate over the training set multiple times before $w$ and $b$ converge.

Report the updated values for the first five updates in the table below $[1$: $0\mathrm{.}5$ for table $+0\mathrm{.}5$ code marks$]$

$\backslash$table$[[$Pass$,$w$,$b$],[0($random values$),,],[1,,],[2,,],[3,,],[4,,],[5,,]]$

Check the results manually, compute the update in $w$ and $b$ by hand and make sure that you get the same results. Write all steps for the computation here, make sure to use the same random values you got from

the previous question.

$W^{\text{n}\text{e}\text{w}}=W^{\text{o}\text{l}\text{d}}+e{p}^T$

$b^{\text{n}\text{e}\text{w}}=b^{\text{o}\text{l}\text{d}}+e$

Where $e=t-a$

Check if the solution for $w$ and b you got is correct. To do so$,$ for each data point, compute the predicted output $($a$)$ given the updated $w$ and $b.$ Then, compare it with the target output $(t)$ and see if there is an error in prediction or not.