I just started R yesterday, haven't coded at all after I did a little HTML in 6th grade, I'm getting it quick but I have a very specific problem that google or any other forum has already answered (and if they have it isn't easy to find!)
I have 6 columns, they work in pairs (IE one column is independent and the other is dependent on said column). My data is average frequencies produced on a spectrum graph from differently shaped trumpet leadpipes. I broke each of the three leadpipes into ten sections, and took a measurement at the bottom of each section, 5 inches, 0.5 inch increments. I then recorded the spectrogram of my trumpet using each different leadpipe, and got three giant frequency graphs that I manually converted into an excel graph by plotting each frequency and the amplitude of its corresponding peak, then split each of my spectrograms into 10 sections just like my leadpipes by averaging the values in each section (each section ended up being made of 15 individual frequency peaks). So, I have each leadpipe measurement set, and each spectrogram data set which corresponds to a specific leadpipe, and I want to train a network to take the knowledge that "these measurements lead to these frequencies" for each leadpipe and be able to feed it a set of ten measurements and have it output a new frequency data set. It's a lot to learn for a first attempt but it has been fun and I'm pretty close. I think
So far I have figured out how training and predicting generally works for simple linear regressions, and sectioned some of my columns into their own objects, I was thinking that I may need to create an object for each pair at some point in the future for training. I also downloaded keras and caret if I need those.
here's my table: (dim=diameter, "six" "H" and "ft" refer to leadpipe models, avgf= avg.frequency)