I have some basic questions about Pytorch that I have not see answered and for me its very much so needed in order for me to get started. First and foremost I want to use the University of Maryland Global Terrorism Database (GTD) part of the START website which has over 190,000 observations, and approximately 135 dimensions (features). I plan on using one hot encoding for the variables, deleting some features out, and using feature engineering for others. That being said, I have some questions:

With the dataset - do I need to delete out the header columns? Or can I leave those in there?

Once I delete out what I don’t want - lets say I am left with 130 features, including latitude, logitude values, as well as total attacks for numbers, - my understanding is that whatever I input it will output feature wise correct? So I will get 130 predicted values for the new features - what does this mean with regards to my input?

How should I set it up?