i$)$ Now, we would like to build and test a linear classifier for the same example above. Assume we

are using the same feature vector and training examples. However, let us add a bias to the input

vector so that it becomes $[$bias$,$ A$,$ B$],$ where bias $=1.$

i$.$ Sketch the graph of the perceptron and indicate all required features and parameters. Is it

binary perceptron?

ii$.$ Use the classifier to classify the class of the first three data points and check if it is correct.

Assume the weight are set initially to $w_1=\left[\begin{array}{ccc}1& 2& 3\end{array}\right],w_{+0}=\left[\begin{array}{ccc}3& 2& 1\end{array}\right],w_{-0}=\left[\begin{array}{ccc}2& 1& 3\end{array}\right]$

iii. Perform a weight update for the ii$.$ Part.

iv$.$ In binary perceptron, what assumption we considered to perfectly classify the data?

v$.$ Perceptron is a linear classifier, hence, cannot classify data that are not linearly separable,

what modification to the perceptron could enable it to overcome this matter $($for both

binary and multiclass perceptron$).$ Give another benefit for this modification.

vi$.$ In binary perceptron, suppose that the decision is wrong hence we have to

change of the weight vector, which of the representations below describe that

considering the initial case.

vii. Which color of the following sketch represents the decision boundary vector:

viii. According to the following graph, at what iteration the training level must stop?