[photon_off]

This was extremely helpful, thank you!

Perhaps you can help me with some follow-up questions:

- Let's imagine a standard excel sheet (a 2d array), with columns "A", "B", "C", "D" ... "Z".

1) Let's say I want to create a model, that takes this input: A single row where all columns have values, except for some (random) columns. I want it to autopopulate those columns with values that are most probable according to the data I trained it with. That is, _every_ column is both a "feature" and a "target". Is this possible?

2) Can I train the model by telling it this: "This input should DEFINITELY NOT confuse you" -- that is, can I "weigh" the inputs (or do I just put in more of them?)

[tlack]

1. You can predict multiple values, but it still has to be trained on target values for those features. So, you could predict "D", "P", and "Z", but not any at random - you'd have to design it that way. Look into "multidimensional regression"

2. That seems logical -- supplying a "confidence level" with the training data itself -- but I haven't heard of it, and can't seem to find anything on the search engines.