I would make a stronger statement: numerical algorithms (computational continuous mathematics) are seriously neglected in favor of combinatorial algorithms (computational discrete mathematics).

Take any undergrad algorithms class: 99% of it is combinatorial. There's an equal (probably larger) parallel universe of numerical thinking and numerical algorithms which is either not taught, or taught as a secondary class under the names of 'numerical analysis', 'scientific computing', and so on. None of stacks, queues, trees, and graphs are going to help you out when you have to discretize a differential equation or solve the resulting linear system with correctly enforcing the boundary conditions. You could ace an algorithms class and not have a clue how to begin tackling those kinds of problems.

This is especially relevant today when machine learning is heavily dependent on computational linear algebra and other numerics, and a typical CS student is not trained for that.

Interesting. I agree, though linear programming does get a spot almost all the way back in "Introduction to Algorithms", which I believe is a pretty common standard text for CS students (one of the authors is an IEOR professor at Columbia). My guess is that most algorithms courses don't make it all that far back. I wasn't a CS student - I was a math major and did an MS in IEOR, and I was a little surprised to see this topic at the end of this book when I was doing some self-study in algorithms. It seems very different, and I hadn't thought of it as an algorithm before, but of course it is, just one that is heavily based on linear algebra.

LP and NLP do seem to be treated as a niche subject, something that would be taught in an Industrial Engineering (or barring that, a math) class rather than as a core CS topic.

Interesting that you mentioned excel - I agree that this is a good way to learn and play around with LP, but normally I'd have a good python library to point to. I used a python library a while back to do some homework problems, and I wouldn't' be surprised if it has improved considerably since then. Still, interesting that LP and NLP aren't part of scikit-learn (http://scikit-learn.org), which covers classification, regression, clustering, dimensionality reduction, model selection, and preprocessing, but not optimization per se, even though non-linear optimization is a key bit of mathematics behind many of these algorithms (regression that uses a cost function almost certainly uses some kind of NLP, right?). It really is notable that LP is not part of the main packages, front and center.

If you go on kaggle or other data science sites, linear and non-linear optimization don't seem to get much attention, either. Wonder why that is. Anyway, interesting topic.

We in cs theory are constantly emphasizing the universal nature of LPs and their generalizations :)

It is written in LaTeX. It even says so in the first few pages.