John Honaker

Montani Semper Liberi, but Buckeye Nation's alright. Statistics PhD Candidate at OSU. Data Scientist at Nationwide.

[John Honaker @MisterMentat]

@cutezu_ *cough* Group actions of vectors on points *cough*

More seriously, I completely agree with all your “missing semester” points!

[John Honaker @MisterMentat]

@jludwig86 @PhDemetri Very mildly “high dimensional” Gaussians would like a word with you.

[John Honaker @MisterMentat]

@matvil @allison_horst Oh man, it’s been so long since I’ve read C&B. It’s probably a big influence on why I think about them that way then!

[John Honaker @MisterMentat]

@dweitzel Thanks! But… How did you know that was me? Search my name on Twitter?

[John Honaker @MisterMentat]

@f2harrell @statsepi Put this in your pipe:

(\(x) { savepoint <<- x; x })()

Smoking it is optional.

[John Honaker @MisterMentat]

@PhDemetri But a careful analysis of your model afterward? Nah. *sigh*

[John Honaker @MisterMentat]

@lexi_lambda As someone that's been using Racket/Scheme for a little over a year that's been learning Haskell for the last month, I think it comes down to two things: the explicit precedence (yay parens) in Scheme, and the density of non-obvious operators.

[John Honaker @MisterMentat]

@betanalpha Universal approximator

[John Honaker @MisterMentat]

@tjmahr This is a really good example for the dangers of defaults in analysis software. Sometimes I think we should all be forced to specify all options at least one time per project. In practice it'd just lead to more copy and pasting from SO though.

[John Honaker @MisterMentat]

@AlexanderKatt This is not a confident guess, but is it "logical implication"? If it is, the second image is clearer; if it's not, I have no idea!

[John Honaker @MisterMentat]

This was a really interesting paper. I’ve always had this (maybe?) controversial opinion that neural networks are really good in domains where generalization is essentially overfitting. It explains why “data augmentation” helps them generalize.

[Miles Cranmer @MilesCranmer]

John von Neumann: "with four parameters I can fit an elephant"

Meanwhile, this paper https://t.co/n15oCHoK3U: "How to fit any dataset with a single parameter"

Here's a function with a *single* parameter. Even worse: it's differentiable and continuous!

[John Honaker @MisterMentat]

@sam_power_825 Hey, that’s my department. Matt is a super awesome guy!

[John Honaker @MisterMentat]

I don’t think I’ve ever seen a better description of grad school.

[Sean J. Taylor @seanjtaylor]

@PhDemetri Grad school is just five years of training you to answer any question with “it depends”

[John Honaker @MisterMentat]

@dan_p_simpson @MikeLwrnc @zerdeve Any favorites, in particular?

[John Honaker @MisterMentat]

@PhDemetri I once had a med student ask me, “What do you guys even study? We covered everything in statistics in our two week public health module.”

[John Honaker @MisterMentat]

@PhDemetri Oh man, really? I do this to my coworkers, but it’s because I don’t want to interrupt their thinking time, and I usually have an open schedule. Do they secretly hate me?!

[John Honaker @MisterMentat]

@ChelseaParlett I asked for a large bag of pepperoni to make pizza. They didn’t have that size, so they gave me *6* regular sized bags of pepperoni. The large was maybe 4oz bigger...

[John Honaker @MisterMentat]

@PhDemetri @statacake A slight alteration of Fried Liver called the Lolli attack Nxd5 followed by d4 is my go to. It’s razor sharp if black deviates from their best line.

[John Honaker @MisterMentat]

I never thought sulfuric acid would trigger a nostalgia wave.