Question 1 –
Consider the following multi-layer perceptron.
Figure 1: Two layer multi-layer perceptron
For the set of input data
X
calculate the output vector y∈R6. You can assume that all activation functions ϕ(ν) are rectified linear units (ReLU), where
Question 2 –
For the following two-dimensional data sets, do the following:
- For each data set, design two different neural networks by hand which can discriminate between the classes (each class indicated by a different color).
- Code your solutions in python and verify that the network learns a decision boundarysimilar to what you designed. For this section, you should include plots showing the data and the corresponding decision boundaries learned by the networks.
Figure 2: Data set 1
Figure 3: Data set 2
Figure 4: Data set 3
