Hacker News new | ask | show | jobs
by graycat 1621 days ago
Part II

If you do much with computer graphics you will encounter matrix theory. That takes you into linear algebra; next to calculus it is likely the most useful math. Evidence: There are a lot of downloads of LINPACK.

Can start a course in linear algebra by considering solving several equations in several unknowns. The standard technique is Gauss elimination, and can program that in about one page of code. Linear algebra is a good start on curve fitting in statistics and the math of quantum mechanics.

If you want to understand more about cryptography and error correcting codes, you should study abstract algebra. Here I would suggest that you actually take a course (a) to help you get through that quite different world of thought and (b) especially to learn how to write proofs. And for (b), take a course where the prof is really good and also carefully reads and comments on your proofs. Abstract algebra is the easy place to learn to write proofs.

Can get more guidance on how to learn math at

https://news.ycombinator.com/item?id=28215105

Somehow long, maybe still, knowledge of both math and computing can be welcome and lucrative in parts of US national security. That was the case early in my career when my annual salary was 6+ times the cost of a new high end Camaro.

Soon FedEx had what their founder, COB, CEO called their "most important problem" -- fleet scheduling. The BoD was concerned, and crucial funding was at risk. I typed furiously, wrote some software, the output "solved" the problem, enabled the funding, and saved FedEx. There, sure, needed to calculate great circle distances so used the law of cosines for spherical triangles -- solid geometry can be good stuff! Also had to handle wind vectors -- linear algebra can be powerful stuff. Then I went off to do much more, integer linear programming set covering where can discover much of the motivation for currently the most important problem in computer science, P versus NP.

Later the BoD wanted some revenue projections. I did a little with some calculus and got a nice answer. Long story short, that work saved FedEx a second time.

For another long story -- I needed to be better at office politics -- I just missed out on some FedEx stock that should be worth ~$500 million now.

The US Navy was collecting ocean wave data at sea, and I was in a software house bidding on writing some software to analyze the data. One customer engineer wanted (a) to know the power spectrum of the ocean waves (that is, what frequencies have the power) and, then, (b) to generate synthetic, random ocean waves with that power spectrum. I quickly read a book by Blackman and Tukey, typed in some software, showed the engineer the results on how to find the power spectra (with an important point about handling low frequencies) and how to generate the synthetic waves, and our company got "sole source" on the software work.

Later at IBM's Watson research lab, we were doing AI for monitoring of server farms and networks. I thought of another way, for some of the monitoring much more powerful than the AI, based on some original math, and published the results.

Net, some math, especially through calculus and linear algebra, can at times be an important career advantage. For more, get good with probability theory, if you can, the version based on the subject measure theory. Then learn some about stochastic processes. E.g., once the US Navy wanted an evaluation of the survivability of the US SSBN (missile firing submarines) fleet under a special scenario of global nuclear war limited to sea -- in two weeks. From some old work by B. Koopman, I saw a continuous time, discrete state space Markov process subordinated to a Poisson process, wrote some software, and was done on time. My work got reviewed by a famous mathematician, and he questioned how my software could "fathom the enormous state space". I answered, at each time, the number of SSBNs surviving is a real valued random variable. It is positive and not greater than the number of submarines to begin with so is bounded and has an expectation and a finite variance. Then the law of large numbers applies. So, generate 500 independent sample paths, average them, and get the expectation "within a gnat's ass nearly all the time". He agreed. I passed the review!

If you go for a Ph.D., then understand that, in the US, academic positions at the better universities are about three things, research, research, and research, especially because that leads to grant money. The operational definition of research is that it got published in a peer-reviewed journal. If you publish, say, 3 papers a year, then likely people will stay off your case and you will likely make progress to tenure. People making the promotion and/or funding decisions will rarely look at the papers and, instead, just count them. Papers that result in prizes are usually quite powerful for a career. Generally, though, academics is not very promising for providing a good standard of living and good financial security for you and your family and these days can't hope to compete with what is available in computing, the Internet, etc.

Then the math? It can be an advantage. The "advantage" can have you push ahead, maybe by a little or a lot, useful technology, economic productivity, and civilization. Such progress happens, actually fairly regularly, but is rarely easy. So, if want to push civilization ahead, (a) don't expect that the work will be easy but (b) math can be one of the most powerful advantages.

Now you know some of what I wish I'd known at the beginning of my career. I want a do-over -- where can I apply?