Hacker News Reader

Comment by rossdavidh | Hacker News Reader

True, but the usefulness of calculus to a working developer in ML is pretty marginal. The difference between different ML algorithms for hill climbing or gradients, is orders of magnitude less than the effect of having the right training data, formatted the right way. Statistics or data science is far more applicable to nearly any field on real-world programming.

But, you know, like Latin in the 19th century was always still useful to one's education, calculus is still useful. It is also something a lot more people know how to teach, than know how to teach statistics (or other more useful topics). I think the latter is the primary reason it remains central to most engineering programs.

View on HN

Agree. Statistics > calculus for the software engineers. Statistics > calculus for most people doing research as their papers would need to interpret t-scores, z-scores, confidence intervals etc. I don't understand this fetish about calculus.

Teaching only statistics and no calculus is how you end up with people such as Tai reinventing high-school calculus and attempting to use statistical methods to validate their “method”.

https://math.berkeley.edu/~ehallman/math1B/TaisMethod.pdf

Oof. Tough look to name the model after yourself. Why is this cited so many times?

Well, you need at least some understanding of calculus to meaningfully understand statistics, don't you? A basic ability to intuit about integrals and derivatives seems like table stakes.

I disagree with this. Being able to read and understand the math in the paper of the algorithm you are implementing is useful, and calculus is common in ML papers.

You may have some short term success without understanding the algorithms at all, but as the field changes and you are no longer in school, being able to keep up at least somewhat with papers is very useful.

I agree that the day to day is mostly about formatting data though!

Well I took calculus, and I did a Master's thesis (much smaller than Ph.D., obviously) on neural networks, and I didn't find much application. Plus, the vast majority of work on ML is not going to be taking an ML paper and implementing it in code. It's going to be transforming this raw data into a format that the existing ML library can accept as input, selecting which cases are useful for training (e.g. making sure each important sub-case is adequately represented), and other things surrounding the data and how it is fed into the (already existing) ML library. Perhaps also playing with the options of the ML library, as to what kind of model you build.

But 99% of the alterations you can make to an ML library, will not make nearly as much difference as what data you feed into it and how. If it's the right data, many ML models will work, and if it's not the right data, none of them will. But regardless, none of this requires, or even really benefits from, calculus.

I agree with all of this, but ML will evolve a lot over the next 30 years of your career. Being able to read the papers as the field evolves is useful, and many of the papers (especially the ones that shift the field) will assume knowledge of calculus.

> Latin in the 19th century was always still useful to one's education,

Was it? Or was it perpetuated by a community that happened to already know it and so they leveraged it?

Latin is what almost the entire Western academic literature prior to the 19th century was written in. You may or may not know this, but Google Translate didn't exist. So, it was essential to anyone undertaking academic studies at the time (in the West) to read Latin, though not to write it. Nowadays English holds that position, despite Google Translate.

Yes, because learning Latin - as a written language - will put your vocabulary far ahead of those who don't. My high school had a general elective focused just on learning to use Greek and Latin to enhance your vocabulary, it was essentially a free SAT prep course for the language section.

So many terms can be quickly understood if you understand Latin prefixes and suffixes, and the better you understand Latin the better you'll understand its use in any of the modern Western languages.

I suspect the answer is "both".

Bingo.

Reply on news.ycombinator.com