You are on page 1of 1

Problem-1 This will help you to understand how a multi-class classifier using perceptron works.

Note that a perceptron is a unit which computes ∑𝑖 𝑤𝑖 𝑥𝑖 as its output (for i=0, the input x0 is equal
to 1 as we discussed in class). Consider three such perceptrons below, which respectively
correspond to classes A, B, and C. You must be observing that for the three perceptrons shown,
their bias weights or w0 are given as 0.8, 0.1 and -0.6 respectively.

(a) Write down the equations of the decision boundary planes for the three perceptrons.
(b) Let us say that for a given input x, the perceptron with the highest value of ∑𝑖 𝑤𝑖 𝑥𝑖 is the
prediction of the group. Then for x = (1, 1, 0) as the input to each of the three groups,
which class is the prediction of the group?

Note my input data in this case is a 3-dimensional one, by convention in books you will see that vectors
are always written in bold, in exams you must write the vectors like 𝒙 ⃗ , ) therefore all the decision
boundaries here will be 3-D plane. In general for N-dimensional input vector or if you are
considering N- number of independent features in your dataset, a linear classifier will use a
hyperplane in N- dimensional Euclidean space. Hope after solving this problem your idea of Linear
Classifiers gets clear. We will discuss Non-Linear Classifiers later after I teach Kernel functions.

Problem 2 Which of the following datasets are linearly separable?

Comment on the dataset (ii) Do you find anything abnormal?

You might also like