Good Day,
In the course 1 we have only dealt with multiple linear regression of the form w1x1+w2x2…+b
In course 2 we have dealt to fit the neural network with polynomial regression of form like
w1x1+w2x2^2…
Could you please explain why is this distinction and use cases? And also uses cases of linear and non linear models and what are they exactly
The difference is in the complexity of the model.
In training and making predictions, they’re really the same. They differ only in how you compute the features x when you create the dataset.
Once you create a feature like x_2^2, it’s simply a constant, and it is handled exactly like how it was in Course 1.
Hello, Thank you for the reply.
How do I decide upon to use a multiple or polynomial regression model which brings me to point of choosing a linear and non linear models. So, I want to understand linear and non linear models and how they affect in choosing the regression model between a neural network and machine learning model
The way I look at it, there really aren’t any non-linear models. There are just linear models that are constructed using additional features from non-linear processes (like polynomial terms).
The way you decide is by iterating on the model design, adding just enough complexity to give good-enough results.
There is a lot of art in this science.