I really appreciate his sharing his notes that translate fairly complex math into something like an intuition. It fits in a helpful (but sparsely populated!) niche between academic papers and tensorflow-for-software-engineers tutorials.
Yep. Second formulae and I'm out of my league. Is it just me that gets frustrated when things aren't explained? What do I have to study to be able to read this
θ∗,ψ∗←argminθ,ψg(θ,ψ)
With no prior explanation and say, yep, now that you put it that way, it's true?
Notation is notation. I think it's hard to come up with a single guide because a lot of conventions depend on the area.
In the context of optimization, a star refers to the value of the variable that minimizes (or maximizes) a function you're interested in.
You will also see a hat over a variable (https://en.wikipedia.org/wiki/Hat_operator#Estimated_value) sometimes (from statistics) which is an estimate of some quantity from data (and that estimate may be obtained by minimizing a function - solving an optimization problem)
I studied post-grad ML and I can say you really don't need to go this deep to use ML - check out http://fast.ai for a code-first practical crash course :)
The discussion 3 months ago on "Part 2" mentioned a re-launch on the "Part 1" content ("redoing the whole of part 1 starting in October"); how's that working out? I wish the tedious maintenance (keeping instructions updated with the latest versions of the software, etc.) was a higher priority there.
As someone more at home in statistics than ML/CS, it's funny that I find this easier to understand than the more 'true' cookbook style postings. The real 'grok' for me would be this kind of mathematical intuition, combined with real data examples. I always learn by just trying to recreate models in books, down to the algebra. If you can feel it and inspect it, you'll get it.
13 comments
[ 1564 ms ] story [ 1563 ms ] threadI found a great article[1] that at least helped me build a simplistic understanding as a modern developer.
[1]. https://simpleprogrammer.com/2017/10/02/machine-learning-mod...
But yeah...It's definitely not a beginner ML cookbook.
θ∗,ψ∗←argminθ,ψg(θ,ψ)
With no prior explanation and say, yep, now that you put it that way, it's true?
In the context of optimization, a star refers to the value of the variable that minimizes (or maximizes) a function you're interested in.
You will also see a hat over a variable (https://en.wikipedia.org/wiki/Hat_operator#Estimated_value) sometimes (from statistics) which is an estimate of some quantity from data (and that estimate may be obtained by minimizing a function - solving an optimization problem)
By the way, this is the first result when googling argmin: https://en.wikipedia.org/wiki/Arg_max
It takes a function g, and returns the values of it's arguments (theta and psi) that minimize the function g.
https://news.ycombinator.com/item?id=14887414
Apparently there's been a bit of "broken links and poorly organized content". If you decide to try it out, don't miss http://wiki.fast.ai/
PS. Ask HN: Best way to get started with AI? | https://news.ycombinator.com/item?id=15689399 (3 days ago)