I’m not so familiar with VirtualBox but I’ll try to follow those instructions. I suggest @carette do as well.

Will do - inserted into this afternoon’s “to do” list.

Meanwhile, I’m trying to address the reviewer comment > The diagrams are hard to read. It is not easy to see which curve represents which system. for GmmGibbsAccuracyPlot.rkt. Instead of enlarging the plots, I think it would help more to add markers to the legends (i.e., overlay point and line legends). To do that, I think I want to override flatten-legend-entries (https://github.com/racket/plot/blob/8dcfd7745e2595b8d517fd8cd7c59510efab84a9/plot-lib/plot/private/no-gui/utils.rkt) to coalesce legend entries with the same label (https://github.com/racket/plot/blob/8dcfd7745e2595b8d517fd8cd7c59510efab84a9/plot-lib/plot/private/common/types.rkt). Does @samth or @rjnw know how to override a library function like that? Or have any other advice how to achieve the desired legend effect?

You can’t override library functions

So like. What do I do

How do I fork a library, say

I mean if you can get hakaru-benchmarks to use its own copy of plot then I can do the rest

(I was hoping for something like overridden-collects in handin)

I guess my first instinct is that not doing this seems like a better idea than forking plot

I also don’t understand what you’re suggesting for the legend

My sense is that the color pallette is not good and picking different ones would help

I changed one line in GmmGibbsAccuracyPlot.rkt to get this:

So I want 4 legend entries, 3 of which have both a line and a point marker

I tried doing what @ccshan is suggesting buy couldn’t find a way. One way would be to manually edit the df through inkscape or something else.

And place those markers on the line

I guess the next thing I’m going to try is to manipulate the curried functions returned by points and lines to merge the legends, possibly using mutable state.

Ok, I’ve downloaded that .ova file from Google Drive. Where are the instructions I should follow?

I know nothing about “virtual box” and “appliances”. [Making me a good beta tester]

you should load the file in virtual box, which you can get from http://virtualbox.org\|virtualbox.org

(note that the artifact reviewers will be familar with starting VMs in virtual box)

Working on it.

Do I need to install the networking feature of virtual box?

you shouldn’t need to

Good - as I don’t want it to mess up my VPN. I believe I installed this software N years ago, and it did.

Ok, I managed to install it, both VirtualBox and hakaru-ubuntu inside it. When I try to start it, I get the error Not in a hypervisor partition (HVP=0) (VERR_NEM_NOT_AVAILABLE). VT-x is disabled in the BIOS for all CPU modes (VERR_VMX_MSR_ALL_VMX_DISABLED).

I guess I need to reboot and enable that. Hate rebooting when I have a gazillion tabs open, once in a while, they all get lost!

Is that success or failure?

[Also, did you see the DM I sent you?]

There’s something very reminiscent of the histogram transformation in this PLDI’19 paper https://www.cs.toronto.edu/~azadeh/resources/papers/pldi19.pdf (paper was spotted by Jeremy Gibbons as it uses monoid homomorphisms).

Well it keeps printing new messages like Doing iter: 0 / 1 and 'gmm' is up to date so maybe it’s success?

I saw something go by but I can’t find it anymore. Where’s this DM?

Would you like to write some related-work sentences about this paper? Also the GLORE LER paper.

I successfully added point markers to line legends. Now I want to do this for all the plots. But some data to plot seems missing from hakaru-benchmarks: the following commands from Makefile fail: racket NaiveBayesAccuracy.rkt racket NaiveBayesLiklihood.rkt racket LdaLikelihoodPlot.rkt --y-min -4200000 --y-max -4400000 50 ldalikelihood-50.pdf racket LdaLikelihoodPlot.rkt --y-min -4500000 --y-max -4700000 100 ldalikelihood-100.pdf @rjnw If you have the data to plot handy, would you please add them to the repository? Also please let me know if I’m not using the right commands.

If you’re using the Slack app, then under Channels there’s another menu called Direct Messages. Look under my name.

It

’s not really published yet, I don’t think we’re obligated to say anything about these.

The problem is I don’t see your name.

I think obligation is not the right framework to think about this

Odd. Let me try sending something again.

Otherwise, what I’m looking for is a better translation of item 214 on the attached menu https://static1.squarespace.com/static/538caaffe4b0371e6dadb904/570efbba356fb0dd1d5f04bc/598731a5a803bb82a16136d0/1502032320497/MenuHamilton_2017-03.jpg?format=1000w

Indeed. Too long a ‘to do’ list has given me a bad perspective.

I need to read it carefully to understand if it goes beyond “reminiscent” and is truly related. And, if it is, I will write about it.

unfortunately I didn’t push that data on github because it pretty big (hundreds of MB).