Finished Discrete Optimization and why it matched my coding style

Whew, done with the class! Was a lot of fun, because it matched the kinds of problems I love to solve and have made a career of solving. I program iteratively, which is the way I solved the classwork. I still need to watch the videos, I mostly just did the assignments because they were fun.

Two things I took away from the class that I always knew but it reinforced: data analysis and visualization are keys, and write functions that take a solution as input and return a slightly better solution, repeat. There are limits to these ideas, algorithms have a limit to what they can do, ie they get trapped in local minima. Then one has to have functions that pop the solution to another valley to get better values. There is an art to this... or is it a science. :)

I also talked about what I bring to the table when I tell people I do perf work for a living - I make code faster, but usually by changing the algorithm or by coding to the data. Rarely do I do assembly any more and I really miss that at times. CUDA I love because it is like programming in assembly, in that one has to keep track of the hardware states to get optimal performance. Though the best increases come from algorithm changes, so a language that allows fast prototyping is my preference as a first draft. Programming hardware for optimal results is addictive/thrilling for me.

And I mention a dream and alien tech again, though in passing. Dreams that seem to last thousands of years... I hate them, because I don't feel rested, but I also like them, because I've seen entire civilizations rise and fall, and advanced tech, etc.