Greetings one and all,
I am beginning to learn Mathematica for purposes of machine learning (had some basic experience using Tensorflow previously). I understand that the Predict() function is a high level tool for generating a regression-like prediction using a specified method.
I have a couple of questions if someone could help me clarify:
1) Are different input types handled automatically? By this I mean, could I simply feed numerical and categorical variables to the Predict() function? Previously, I had to perform one-hot encoding for categorical variables, and I noticed Mathematica has the NetEncoder() function. Is it mandatory here? In a video (https://www.youtube.com/watch?v=lczqhcnVQ8c) around the 27:00 mark, it states that Mathematica are trying to develop multiple output types, but what about inputs?
2) Do I need to normalize the variables?
3) I am assuming that in the Predict() function, if the Neural Network method is specified, it is only a shallow network. What is the approach for me to follow in Mathematica to build a deep learning network?
Thank you and have a great day. Looking forward to hearing your advices.
Did you try carefully studying documentation that has tones of examples for probably all your questions? For example, the last one 3) is clearly answered in examples of NetTrain and related guide:
Did you actually read through all examples in Predict docs to get a sense?