on E&M mostly because I never took the undergraduate class in it. I also didn't take thermodynamics but I studied hard for it and I think I learned it pretty well.
I went to a conference in Australia last year. There were proceedings published but you had to type your contribution up in LaTeX. There was an older amateur there who'd put up a poster on the subject of the Carnot cycle for photons. He didn't know LaTeX so I volunteered to type up his notes for him. I did this to be nice, but it turned out to be a lot of work and gave me an easy way of learning some stuff about thermodynamics. And his paper was published with the proceedings:
"Does Electromagnetic Radiation Generate Entropy? The Carnot Cycle Revisited"http://scitation.aip.org/dbt/dbt.jsp?KE ... 46&Issue=1
His paper was basically a corrected calculation for something that was first most of a century or so ago, as far as I could tell. I think that a good way to learn more physics is to get involved with other people working on things, especially if it seems like an interesting idea in an area about which you know little.
It seems that it's easier to teach yourself thermo than it is to teach yourself E&M as the math is easier. On the other hand, I've had a dozen classes in probability and statistics along with a year and a half of graduate statistical mechanics and I'm strong in that area. And I don't have any great problems in atomic physics.