Week 3 Programming Assignment: Intuition behind non-linear borders between classes

Course Q&A Deep Learning Specialization Neural Networks and Deep Learning
1

Hello everyone!

I would like to get some intuition behind why when we compare network models with 1,2,3,4,5,20 and 50 neurons in the 1st hidden layer, 1 and 2 neuron models are only generating linear borders between classes, and then starting with the 3 neuron model it is able to create smth more complex.

image
image700×537 58.9 KB

Thank you in advance for your answer!

Best regards,
Anton

1 Like
2

Hi @Dammitcarl,

Adding in more neurone help your model to learn the training data better, hence, come up with a better decision boundary.

Cheers,
Mubsi