Polynomial z

All examples and practice labs use a linear expression for z = w * x + b. Would it be useful to leverage polynomial variations of z? Could you share examples?

Hi ljb,

Can you explain in what you want to leverage polynomial features of z? I mean in what pretext you are asking, if it is related different features related to variables x’s which gives a dependent variable z??

I don’t have any specific use case in mind. In fact, that’s what I am trying to figure out: When would this be useful?
We used both linear and polynomial z functions for linear and logistic regression in the previous course of the specialization (i.e. Supervised Machine Learning). So I am wondering whether this would also apply to neural networks.

Hi ljb,

Polynomial features are a type of feature engineering. For example the creation of new input features based on the existing features. The degree of the polynomial is used to control the number of features added. So we do use in neural network too but relation part would not be always linear as in neural network, we create model with nonlinear complexity and in such cases features are either added, modified, labelled to fit into polynomial model fit which is then trained.

Thanks. Just to clarify, you are saying we could use something like z = w1 * x + w2 * x2, right? Would you have any actual example to share? Maybe a scientific paper or tutorial.


can I know if you have done DLS specialisation?

your reply tells you still didn’t understand my reply completely. in supervised learning algorithm it is always a linear and logistic regression where it might have a relation like

but as you go in advanced learning algorithm or unsupervised algorithm, the relation cannot be always linear as we create different model algorithm based on different feature which can have nonlinear relationship too. So the equation you share will not always fit. you will understand when you do 3 course of MLS

For now, I only attended the previous course of the specilization. I realized I made a typo. I meant: z = z = w1 * x + w2 * x^2.