My name is Jeremy Kun. I’m currently an engineer at Google. I earned a PhD in mathematics from the University of Illinois at Chicago, where my advisor was Lev Reyzin. Here’s a long thing I wrote about my graduate school experience. I did my undergraduate degree at Cal Poly San Luis Obispo in mathematics and computer science.

I have a thorough background in computer science, but my most exciting experiences there always stemmed from elegant (and uncoincidentally mathematical) solutions to programming problems. This blog is a presentation of the interesting solutions I come across, and an exploration of the deeper mathematical ideas therein. Most often this means exploring the mathematical structure of a problem to lubricate the cogs of algorithm design. In seldom cases, this involves using programs to reason about mathematical theory.

I try to reserve this blog for technical material, and post articles for a general audience on my medium blog (and others, see this historical archive).

### Want to support Math ∩ Programming?

Check out the support options.

### For potential employers or research partners

Think of this blog as a three-fold portfolio: code, technical writing, and distillations of ideas from various mathematical and software fields. Most of the work on this blog has followed the recipe:

- Stumble upon a topic that sounds fascinating, or become fed up with my own ignorance of some area of mathematics or programming, or come up with an original question or idea.
- Read books, papers, blogs, and watch lectures online about the topic.
- Come up with a small software project that uses the idea in a nontrivial way.
- Write a main content post explaining and demonstrating the project.
- Write a primer post explaining the background mathematics.
- Look for ways to expand, generalize, or relate the idea to projects past and forthcoming.

“Main Content” posts contain software, while “Primers” are usually math-only. All source code used in making posts is available at my Github page.

Here is a list of the languages used in some nontrivial capacity (using nontrivial features of the language) on this blog so far:

- Python
- Mathematica
- Javascript
- Java
- C++
- Racket
- Standard ML
- Processing

Here are some other of my online profiles and pages related to this blog:

- Github
- Academic CV
- Twitter (personal)
- Twitter (blog feed)
- Medium
- Facebook group (unmonitored)

Math ∩ Programming by Jeremy Kun is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported License.

You’ve been doing a great job here! Congrats for your initiative.

LikeLike

Thanks! Now that I’m in graduate school I’m finding less and less time to work on it. Luckily next semester I’m gearing my courses toward topics that are quite blogworthy.

LikeLike

Thank you very much for your blog. I enjoy it. Bookmarked

LikeLike

You have an excellent blog. Thank you very much for organizing and presenting all this information.

LikeLike

Your Have amazing blog. thank you ; I enjoy it

LikeLike

Excellent work ! Thanks for all the posts .. Please keep it up ..

LikeLike

Nice blog. Simple and interesting. This stuff is incredible.

LikeLike

Jeremy – I enjoy reading your Math Blog, and would like to respectfully ask a question.

I was wondering if you had ever done any research on buying

mathematical algorithms vs. programming them yourself? Especially for complicated

mathematical subroutines, is it cost effective to subscribe to an algorithm library

(like http://www.nag.com) for about $3000 per year, or let your programmers do all the work?

Have you ever offered an opinion on this subject (or anyone else)?

Thanks, Steve Chastain

LikeLike

That is a good question, especially because I’ve never heard of the concept of ‘buying algorithms’ being implemented before. In particular, an algorithm isn’t copyrightable material, so what they’re really selling you is either 1. a particular implementation with special features, or 2. their interface or service in deployment, support, and maintenance of said implementations. That’s how companies like Wolfram Research get away with charging for their wonderful Mathematica language.

Of course, I’m going to be critical because I study the mathematical background and implement the algorithms for fun, and I wish all my friends and their mothers did too. But I do believe that the methods which have stood the industry test of time will always be open source, and before you go to pay for a software implementation, it’s cost-effective to attempt to integrate such a library into your own code. I think, at the very least, open source libraries suffer orders-of-magnitude more scrutiny than something like NAG, and hence are bound to have fewer bugs.

To answer your question more directly: if you can, don’t pay for it and don’t implement it either. The optimized versions of these basic techniques have gone through decades of scrutiny to arrive at their present states, and you shouldn’t be reinventing the wheel for a real-world application. Instead, stand on the shoulders of giants, and realize that most giants which are accepted as ‘useful to stand on’ are free.

Looking at NAG’s webesite, it looks like the majority of the algorithms they offer are as such. For instance, their C library touts a Fast Fourier Transform, but the numpy library has the same routines, and more (NAG doesn’t support > 3-dimensional FFT). Furthermore, there is no mention of a parallelized version of the FFT, despite their emphasis on high performance computing and the existence of the industry standard Fastest Fourier Transform in the West again being free, open-source, and implemented in all the same languages.

Even more curiously, it seems that a number of their linear algebra routines are taken directly from LAPACK and BLAS, both free and open source linear algebra libraries specifically oriented for massive parallelism and used regularly in exascale computing applications. Many of their other listed algorithms seems pretty standard too. Although I don’t have much knowledge in statistical computing, I’m certain that the R programming language is far more capable than anything they have under statistics and nonparametric methods. Similarly, most nonlinear optimization techniques, including their packaged forms as the commonly used Support Vector Machine, are widely available. This reinforces my notion that their main service is support, and not the algorithms themselves.

That being said, I have never seen a black-box approach to mesh generation, and they list that as one of their features. As far as I know (and I currently see this everywhere in my work at LLNL), efficient mesh computation on nontrivial geometric shapes is highly problem-specific, and largely an open problem.

I also think it really depends on the focus of your software team and how cutting-edge the techniques are. If your devs are resistant to read a potentially dense paper on some new technique (and I’ll admit it, most papers are impossible to read), then hiring outside talent would allow your employees to focus on the work they enjoy instead of banging their heads against a wall over what can often be bunk or outdated research. It’s a philosophical debate for whether I should reimplement the algorithm or use a library, but for you it may easily be about team welfare.

Did that answer your question?

LikeLike

Hi Jeremy, I have (or would like to have) a similar background as you. I am finishing up my Bachelors in Mathematics from a local public university, but have a lot of expirience in IT and also in programming various fun mathematical based things for myself. I have long been interested in the “mathematical aspect” one could say of finding solutions or algorithms to problems and the analysis of them.

Anywhozel, I really enjoy your blog (a friend just sent me a link to it), and hope one day to find myself as far along in my education as you.

regards,

LikeLike

You really should compile a lot of this stuff into a book of some kind. I would buy it in a second. So would many others, I am sure.

LikeLiked by 1 person

I would buy it too!

LikeLike

Maybe you will find MathJax (Javascript LaTeX Interpreter) interesting. It makes you able to write Math in both MathML and LaTeX syntax directly in the html. Everything is handled by javascript. Which means that there will be pretty math without having the trouble of generating and uploading images.

LikeLike

Unfortunately I’m not hosting my own page, and wordpress doesn’t allow any javascript to be run on their pages for security reasons. But I have heard MathJax is nice.

LikeLike

Maybe it depends on the version of WP you are using? I read that wordpress.com does support Latex http://en.support.wordpress.com/latex/ and this is a plugin I found for Latex http://wordpress.org/extend/plugins/latex/ that can either use MathJax or Latex image server if you have JS turned off.

Great blog! I studied math for my degree and masters, covering pure, applied and applicable for two years before specializing in pure and finally analysis (especially Banach Algebras). I had also programmed for many years and after masters went into programming business. So! the intersection of math and programming is really cool for me.

PS Did you read about the automatic math proof generator that Tim Gowers and Mohan Ganesalingam programmed whose output is close to human written proofs? Very interesting work http://gowers.wordpress.com/2013/04/14/answers-results-of-polls-and-a-brief-description-of-the-program/

LikeLike

Ive been reading your blog for the past couple of months. Thanks for this is great resource.

Might I suggest adding it to google “currents” would make the reading on a tablet much nice

https://www.google.com/producer/currents

LikeLike

Good call. I never knew this was an opt-in kind of thing. EDIT: Looks like I have to be able to “verify” that I own this website in order to publish on Currents. Since WordPress owns the domain and I don’t have access to the HTML headers, looks like I’m out of luck.

LikeLike

Hello j2 this is a great blog, which I reached via your StackExchange contributions. I just read and much appreciated your hitchhikers guide to mathematics for programmers (joke):

https://jeremykun.com/2013/02/08/why-there-is-no-hitchhikers-guide-to-mathematics-for-programmers/

My PhD is in proof-theory with a cs flavour but that was many years ago; I went on to research in machine-assisted formal reasoning with Burstall and Fourman at LFCS in Edinburgh and then taught CS at Sheffield Uni for a decade or so. I now work as a maths teacher with the Open University and have a very new 1-man band software consultancy business (too new to have a proper website yet). I’m learning and using F# and loving it; I’ve been around the block with a number of FP languages and F# is really up to scratch. [I don’t mean that it is as good as the language “scratch” (http://scratch.mit.edu/) :-] But my favourite platform is Mathematica and I use it daily for teaching, exploring, musical investigations, simulating, designing, accounting, drawing, mapping, testing out new programming language features. I use their browser plugin for interactive/animated content; see

http://fairflow.org/wpOU/wp-content/uploads/2012/06/Lissajous1.cdf and use the slider or animation controls. But I guess you know about this already?

LikeLike

I didn’t mean you know about my little animation: i meant you know about cdf and the plugin!

LikeLike

great blog, do you have a facebook association also?

LikeLike

Hey great blog, I really enjoy your posts!

Any advice you could give to a current undergrad student who likes both Math and Computer Science on how to define possible graduate school research areas? I’m currently doing research at my university but I don’t think that’ll give me a good enough overview of all possible research areas. Thanks in Advanced and agan Great Blog!

LikeLike

It honestly depends on what you like. How much math have you done (in coursework or on your own)? How much CS? What topics are you most interested in?

LikeLike

Perhaps I should also say, my institution (UI Chicago) is looking for more Mathematical Computer Science graduate students. It’s more focused on the mathematical areas of CS (algorithms, data structures, learning theory) than the applied areas (systems, networks, security), but I think it’s a great department. And if you like the engineering aspects more, UI Urbana-Champaign is a great school for that too.

LikeLike

On the Math side I have taken Abstract and Linear Algebra, and am currently taking ODEs. On the CS side I have taken Data Structures, Computer Architecture and am currently taking Databases. I’ve been doing research in Finite Fields for a couple semesters now.

There’s a couple topics that interest me, but since I haven’t had much exposure in them I’m not really sure if I’m completely interested. These topics include: Algorithms, Machine Learning, Theory of Computing and AI.

Any advice on how to expand my knowledge on possible research fields, or finding out exactly what field I like the most is appreciated. Thanks again!

LikeLike

Heck, I say take as many classes as you can in all of these topics. If you’re interested in finite fields and abstract algebra, I’d image you’d enjoy theory of computing and the followup would be computational complexity theory. Algorithms is definitely essential. I personally hate differential equations (and I know nothing about PDEs), and it doesn’t appear to affect my work at all.

Machine learning theory is a very hot field right now, and it sits at the crossroads between probability theory, computational complexity (part of the theory of computing), and game theory. It would be a comfortable middle between theory and applications, if that’s what you like. Two big names in that field are Kearns and Vazirani, and they have a lot of survey papers and books you can get to introduce you to the field. On the other hand, there are loads and loads of interesting topics just in complexity theory, and there’s no shortage of open questions. In fact, a lot of important arguments there rely on facts about finite fields, so you’ll get to see some interesting applications there.

I think if you’re interested in going to graduate school you don’t have to pick a field now. I don’t think anyone is expected to know exactly what they want to do right away, but should know that there’s at least one topic they would enjoy studying. If you want, you can pretend in your application that you know exactly what you want to do, and nobody will hold you to it if you change your mind later. Having this finite fields research under your belt will help you a lot with that. If you are looking at schools and you contact a professor you’re interested in working with (and you can start that as early as you want, if you can manage to understand some of their papers), then they can give you a better idea of what their topic would be like.

LikeLike

A friend of mine sent me a link to your blog. And within a glance I was overwhelmed with the content of your blog. You put a great quality in information, plus detail explanation and demo.

I never take blogging seriously until I come to your site. Maybe I should start blogging interesting things I find, and open that knowledge with other people like you did. Great job, and truly adorable, Jeremy!

LikeLike

Thanks by this useful information, are really interesting some of the topics described in this blog. I hope you continue sharing these pieces of science :). Greetings from México.

LikeLike

Hi Jeremy,

My name is Kevin Gillespie, & I live in Wales. Pleased to meet you. :).

I am now ‘Following’ your Blog. 🙂

LikeLike

Your blog is a fantastic collection of proofs and mathematical experiments, a must-follow for students in Computers-math background.

LikeLike

Hi Jeremy, I really enjoy your blog. What do you use to create equations/formulas and such in your posts?

LikeLike

WordPress supports LaTeX, which is the standard math typesetting language. You can type equations in comments by wrapping in dollar signs and putting “latex” as the first thing inside the dollar sign. see http://en.support.wordpress.com/latex/

LikeLike

Very nice blog. Good job. In fact you are the first academic that I see to take Apple-like approach. To care about the design(call it Art) of the website when dealing with hard science at the same time.

LikeLike

Hi Jeremy, Your blog is the best I ever seen on Math and CS. I wonder if you could share your favourite books on the subjects, especially these below:

Linear algebra

Graph theory (Discrete Math)

Calculus

Probability theory

I a CS professional, and want to fresh all the subjects from basic to advance levels, like you.

LikeLike

– Linear Algebra Done Right – Sheldon Axler

– I have never used a Graph Theory text

– Depends on your goals, but probably Measurement by Paul Lockhart (or Spivak’s Calculus on Manifolds, but Calculus is really not all that good for CS)

– I have never actually read a full Probability Theory text

LikeLike

Thank you for the prompt reply and the advice.

LikeLike

Thanks, from another Jeremy!

I’m a programmer obsessed with signal processing but lacking in formal education, especially in math. I realized linear algebra in particular is something I need to understand, but I was having a hard time figuring out where to start. Your primers gave me an excellent starting point, and from Googling things I didn’t understand I found a wealth of information I didn’t know how to find before.

Slowly but surely I’m starting to get this!

LikeLike

You’re welcome! Yes, linear algebra is extremely important. My favorite book on linear algebra basics is Axler’s “Linear Algebra Done Right.”

LikeLike

I read your series on elliptic curves today, I couldn’t stop, it was such an enjoyable experience just reading the carefully constructed well written explanation of it all. I specially liked the detour on the projective space, made perfect sense. I wonder if you are planning to expand on Montgomery curves and Twisted Edwards curves as I’m trying to learn the details of Bernstein’s Curve25519. Love your work!

Thank you Jeremy!

LikeLike

I was actually unaware of these different forms of curves, and I’m surprised to see how they avoid issues like side-channel attacks! While it may take a bit longer to get to these topics (it looks like you need some algebraic geometry to understand why anyone would think about a twist), I will definitely be looking at them.

LikeLike

Could you pls make an RSS feed…? most of your target audience hardly ever login to twitter, facebook etc…

LikeLike

All wordpress blogs have a built-in RSS feed: https://jeremykun.com/feed/

LikeLike

Hi, I just read your “math class as art class” and “you never did math” posts. Very interesting, and I think we may have similar ideals. Here is a rough description of a math curriculum I’m working on that might add some similar and different perspectives. http://www.effortwise.com/math.html

LikeLike

Hi Jeremy, nice blog! I thought you might want to know that a significant portion of the left hand side of thecontents are cut off when viewed on my Nexus 7.

LikeLike

This blog is AMAZING! Thank you very much!

LikeLike

hello kun, thanks so much for your blog and i personally learned a lot from you. i am wondering have you ever explored expert learning in ML, something similar to bandit learning. thanks

LikeLike

Is there any chance you could make your RSS feed full text rather than just the first few lines?

LikeLike

Hi Jeremy. I just started revisiting analysis with the book “Real Analysis: A Constructive Approach by Mark Bridger” [1] (because constructive mathematics whetted my interest after reading about it a couple of times), and, incidentally, I just stumbled upon a comment of yours on HN that you wished there was an analysis book which emphasizes computation. The introduction of Bridger’s book mentions that it was partly written with computer scientists in mind who want to know about computability and calculability of the reals, so perhaps you will find it interesting. If you already know this book I would appreciate feedback. Thanks for your great blog articles by the way!

[1]: http://www.amazon.com/Real-Analysis-Constructive-Mark-Bridger/dp/111835706X/

LikeLike

The Amazon preview looks a bit strange. Is the typesetting legible in your copy?

LikeLike

That is the preview of the Kindle version and it’s possibly an issue of the Kindle format to HTML conversion. The Kindle versions are seldom well typeset, but I can’t imagine it being that bad. The paperback version is well typeset.

LikeLike

Hi Jeremy

I just wanted to say that I really enjoy reading articles on your blog. They are so easy to understand for a lay person as myself. I especially liked how to take a calculus test. I just wish it was available to me when I was in college taking calculus and differential equations which I at times struggled with. But the good news is that I did pass with satisfactory grades.

I have a question for you. What is the focus of your PhD thesis or research?

LikeLike

I work in a field called “theoretical computer science,” but my work specifically has been in a mix of network science, computational complexity theory, and machine learning.

LikeLike

Hey Jeremy!

I know you write about stuff you’re interested in, so I’d like to interest you in something and maybe see a post on that in the future 🙂

You see, with the advances in machine learning and related the role of floating point numbers has increased. Indeed, in “classical” computer science we mostly dealt with inherently discrete structures (integers or structures like graphs) whereas ML and Probability & Stats introduce continuous numbers to Computer Science. Unfortunately, our floating numbers (IEEE 754) aren’t as good substitute for reals as our integer arithmetic for mathematical integers. Programmers often (and I’m guilty, too) use reals as they were reals, which may lead to problems.

It’s best summarized IMO by a legend of CS, William Kahan who seems to be the only person in the world to be concerned with floating point arithmetic issues. See his talk at Heidelberg Laureate Forum: it’s definitely worth it

http://www.heidelberg-laureate-forum.org/blog/video/lecture-thursday-september-26-william-morton-kahan/

I think this is an important topic and more programmers and researchers should be introduced to it.

Looking forward to your opinion!

LikeLike

Hello from France!

Thank you very much for your nice artitisc mathematical animated gifs!

Jean-Claude

LikeLike

Hi,

Thanks for such a great website. I came across your article on “Where there is no Hitch hikers guide to mathematics for programmers” and I am deeply inspired to restart my journey on mathematics. I have basically always stumbled through it. I’m studying for a masters degree in Machine learning and my choice of course was as a result of my love for programming. However, on getting to class, I am usually lost after the first few mathematical sentences. I was considering dropping out, until i found out some of my class mates feel the same way. It’s only the third week. I’ll be happy to give myself another chance. I wish I found this blog much earlier. Thanks again!

LikeLike

Great blog! Really enjoyed your post on the GI problem and I look forward to your next post on the topic!

LikeLike

This looks great 🙂 I’ve only just stumbled upon this blog by being recommended! Will definitely take some time out as the material covered looks fascinating and just what I’d wanted! Thank you!

LikeLike

You have a great Blog! Full of useful ideas and clear explanations. Thanks a lot for sharing this archive of knowledge.

LikeLike

I love your blog. I was introduced to it about three and a half years ago by a peer of yours from Cal Poly. Since then I return to it every few months, using your blog as an unofficial ‘yard-stick’ for my own mathematical development. When I first visited, I found the topics fascinating but I could barely understand anything. Now I understand enough to use your blog as a guide while embarking upon individual study of the topics you cover here. Thank you, and keep up the great work.

LikeLike

Oh my. I foresee many evenings lost to this blog.

Looking through some of the topics you’ve covered, I think you might be interested in Maurice Herlihy’s (with others) work applying combinatorial topology to proofs in distributed computing. Basically, the state of a distributed computation is a simplicial complex; a protocol may be defined in a way that it represents a transformation from one simplicial complex to another. This allows one to provide impossibility proofs for many distributed problems (e.g., reaching two-party consensus with a protocol in which messages may be lost).

Herlihy delivered a series of lectures on this topic at the Technion: https://www.youtube.com/watch?v=05knu1z3zOg&list=PL0DA9BFB82ACED0AF

and he has a book: “Distributed Computing through Combinatorial Topology”. The book is a text for a class at Brown, slides for the class are available here: http://static.cs.brown.edu/courses/csci2951-s/

Anyway, thanks for this blog.

LikeLike