3

Perceptron Training

Unsolved
Supervised

Difficulty: 7 | Problem written by ankita
Neural networks are composed of small units which are sometimes called neurons or perceptrons. These small units are supervised classifiers of their own.  

Here, our aim is to implement the perceptron training manually using NumPy only.

Given below is the perceptron criteria that is to be minimized to learn the weights:

Ep(w) =  \(\sum_{n\in M}^{}-w^{t}\Phi(x_{n})t_{n}\) 

M: misclassified points.

\(w\)Weights of the model

\(\Phi\)Feature function of x in our case it is identity function \(\Phi\)(x) = x

\(t_{n}\): True labels of training data labeled as +1 and -1 for the two classes

Input:

x: an array of training examples
y: an array of outputs corresponding to each training example (possible values are 0 and 1)
lr: the learning rate for the algorithm
iter: number of iterations the algorithm will perform

Output:
A list of updated weights after every iteration. Do not include the first W that is an array of zeros

Algorithm: 

  • Convert the given y from 0,1 to \(t_{n}\) which is +1 and -1 for the two classes
  • Calculate,  \(w^{T}x\) , then write y_p as an array of +1 for positive value at the corresponding location in the calculated vector and -1 for negative value
  • It can be seen for correct classification \(w^{T}\Phi(x_{n})t_{n}\)  is positive and negative for incorrect classification
  • For incorrectly classified points calculate derivative of Ep(w) wrt to w using the formula given below: Derivative of  Ep(w) wrt to w =  \(\sum_{n\in M}^{}-\Phi(x_{n})t_{n}\), where M is misclassified points
  • Update the weights using the formula: W = W - learning_rate*derivative of the perceptron criterion with respect to W
  • Continue the above step for the number of iterations specified in the function.

Hints:

The first element of W should be wo i.e., if Y = wX + wo then W = [wo, elements of w]
To get the above W, stack a column of ones to X at the beginning of X.

Sample Input:
<class 'list'>
X: [[0.26703489, 0.50235526], [0.18151398, 0.07526641], [0.97665297, 0.09124986]]
<class 'list'>
Y: [0, 1, 1]
<class 'int'>
iter: 5
<class 'float'>
lr: 0.1

Expected Output:
<class 'list'>
[array([[0.8 ], [0.8841833 ], [0.98334837]]), array([[0.6 ], [0.76836661], [0.96669675]]), array([[0.4 ], [0.65254991], [0.95004512]]), array([[0.2 ], [0.53673322], [0.93339349]]), array([[5.55111512e-17], [4.20916525e-01], [9.16741865e-01]])]

This is a premium problem, to view more details of this problem please sign up for MLPro Premium. MLPro premium offers access to actual machine learning and data science interview questions and coding challenges commonly asked at tech companies all over the world

MLPro Premium also allows you to access all our high quality MCQs which are not available on the free tier.

Not able to solve a problem? MLPro premium brings you access to solutions for all problems available on MLPro

Get access to Premium only exclusive educational content available to only Premium users.

Have an issue, the MLPro support team is available 24X7 to Premium users.

This is a premium feature.
To access this and other such features, click on upgrade below.

Log in to post a comment

Comments
Jump to comment-185
mo_venouziou • 2¬†months ago

0

Note the instructions have a couple misprints. The solution initializes W with ones, not zeros, and the dot products should be x dot w, not w.T dot x in order to align properly.

Ready.

Input Test Case

Please enter only one test case at a time
numpy has been already imported as np (import numpy as np)