@pravnar I just saw you commits for newsgroups, is it also possible to store the arrays generated from chad’s script into an input file. This way I can use the same numbers for testing and benchmarking.
Looking at the Augur paper, they also do the compile-time specialization that we do, primarily to drive parallelism choices: https://arxiv.org/pdf/1312.3613.pdf
Also the Augur paper has the kinds of graphs we want to produce
the augurv2 paper also has similar graphs: https://danehuang.github.io/papers/augurv2.pdf
I just pushed updates to benchmarks.cabal so that “stack build” inside runners/hk will produce a getNews executable. It takes two arguments: the path to the newsgroups data, and the path where you want to store three files with extracted vectors (words, docs, topics).
@carette I just pushed a start for the histogram section, but I’m meeting a student right now.
@ccshan Just finished my gscale group meeting. I need to go eat, and then I work on the paper. I’ll make sure to coordinate through here (as well as do frequent check-in/push)
oh nice, thanks
And then Tara needed help with her logic course (which is surprisingly advanced for a 2nd year course!).
I’m definitely in the middle of the histogram section right now
Ok, I’ll work on something else (in the paper).
Such as reordering the Dirichlet section.
Hmm, I don’t seem to have ‘iftex.sty’. Is it standard?
Never mind — texlive-generic-extra.
@ccshan Does it make sense for me to try to change the running example to go to y being a ‘scored input’?
It is not necessarily difficult, but it is a bit dangerous, in that it is easy to make a mistake while changing roughly 20 equations and the accompanying text.
I’m not sure if it’s worthwhile now… Here’s what I suggest:
I push what I have and you read it. Meanwhile I finish this subsection and you take over. The remaining subsection is to describe the implementation of the optimization. The subsection I’m about to finish is about the syntax and semantics of bucket expressions.
Sounds good. Working on notation in the meantime.
Pushed
Will look soon.
In the meantime: there is no explanation of the difference (in the text) between the \vecdots and \vecdotsvec notations. I know what you mean by them, and the subtle difference — but will our readers?
Pushed a variant notation. Looking at your text now.
I pushed my work. Please pull before reading Section 4.1.
Then, your work in Section 4.2 is (I hope) cut out for you.
As for notation, I think using bold is a fine idea, but you don’t seem to have changed the basic \vec macro…
Right - because we have 3 kinds of vectors.
If a basic vector of numbers is \vec{x} then it feels weird for a vector of vectors to not have the arrow. So I suggest \bm{\vec{#1}} for both \vecdots and \vecdotsvec. I don’t remember the precise semantic distinction between \vecdots and \vecdotsvec but I don’t think it’s very important…
That was my feeling as well. I’ll make that change, right after I push the minor tweaks I did to 4.1.
Great. I’m off to eat soon.
Typing rules make everything look more official…
I have to drive home soon (still at the office).
When are we getting self-driving cars?
I can continue to work a bit then, and then from 10:30–15:30 tomorrow are allocated to this paper as well.
I think my next task is to draft Section 2, then I’ll ask you to smooth the transition from Section 2 to Section 3.
If you listen to my colleagues working in automotive and on safety aspects of SE, you don’t really want them too soon…
Yes, I can take section 4.2. And then smoothing sounds good.
I plan to write the introduction in the morning
I am writing a script for plotting curves