Hats off for the Beautiful Course, In your lecture on issues with Linearity, can deep learning be used for regression. In general output of a neuron is between 0 and 1. How does it transform to big number, say salary.
Can we have final output a big number?
Thanks for the kind words.
A regression is just a particular case of neural networks.
You can use unbounded activation functions for the output layer of a regression (so you are sure that you can achieve every real number). This rules out relu and sigmoid. Apart from this you may decide to not use an activation function at all. This will also ensure you can reach any real number.
The 365 Team