1. Implementation of Single Layer Perceptron (SLP) (Total: 10 marks)
Classify the IRIS dataset by using the single layer neural network.
(a) Download iris data from UCI web repository. Click her e
(b) Read the data: input features in one variable and class labels into another in a vector form suitable for a neural network class label representation (2 marks)
(c) Randomly select training and the test set: x% (begin with x=10) data from each class for training and all the rest for testing (2 marks)
(d) Compute training and testing accuracy using SLP for 10 independent simulations and store the results from the individual simulations programmatically in an excel sheet (2 marks)
(e) Compute training and testing accuracy by varying accuracy x (from 10% to 60%) as the following and report overall training and testing accuracy (average over 10 simulations): (2 marks)
1.
Amount of randomly
selected training data
2. Training accuracy (Average over 10 simulations)
3. Testing accuracy (Average over 10 simulations)
10%
20%
30%
40%
50%
60%
(f) Plot a graph keeping column 1 at x-axis and column 2 and 3 (at the same figure) at the y-axis. Use curves of different colors to denote curves for column 2 and 3. (2 marks)
2. Classify the IRIS dataset using a multi-layer perceptron neural network. Repeat the same steps as mentioned in Question-1 using MLP. (10)
