@b3m2a1 Thanks, but I won't be able to contribute much to development, so I'll just hang on until you've got it sorted :)

I hate to DV, esp. answers -- I'd rather just ignore bad ones. But this one got an upvote, although anyone with enough calculus to understand the question should be able to figure out that the integral is divergent (unless somehow I've made a blunder). The question should be closed.

@Silvia Hi, Silvia. Hope you've been doing well! :)

@MichaelE2 Hi there! Long time no see! I guess work is too busy for me to participate the community as I used to, but yes I'm doing great :)

@yode Find somewhere to put your dataset and I'll give it a go :)

Is anyone here experienced with HTML export? I would like to set a special style for inline cells. E.g. make a text cell, press Ctrl-9, enter an expression, exit the inline cell, type more text.

HTML Export has a "ConversionRules" option. This sets how different styles are converted. The inline cell does not have an explicit style, even though it displays specially within notebooks.

So it's not clear if ConversionRules can be used at all.

Perhaps the best way is to pre-process the notebook, identify the inline cells, and give them a style? @Kuba Do you have experience with this?

Sometimes I really do not understand why some questions get upvotes. This is simply a C question that has nothing to do with Mathematica. Let me rephrase: "Mathematica has a function to compute histograms. Now I want to write a C function to compute histograms. How do I do it?" Mentioning "Mathematica" does not make this on-topic ... it is still a pure C question.

I did not flag to migrate to SO because as currently written, it would certainly get downvoted within minutes. The SO crowd wouldn't have the patience, and they're unfortunately not nice. That, however, is not a good enough reason to allow pure-C questions here.

jeepers. why doesn't python have a 'flatten' function softwareengineering.stackexchange.com/a/254676

lambda z: [x for y in z for x in y] is definitely way more principle-of-least-surprise than just flatten(z) 🙄

</pythonbashing>

@Szabolcs I once had to make a quick but customized export from .nb to .html and I went with full custom .nb->.xml->,html exporter because I was not feeling in control of what built-in exporter does, even with ConversionRules etc. It was not very general but enough for the project.

@Kuba What type of XML did you export to? And how did you handle the XML -> HTML conversion?

Does this link load for anyone else? community.wolfram.com/groups/-/m/t/237660 Recently, whenever I use Google to search W|C, then click on search results, I arrive to empty pages. Recently this is always the case. (@VitaliyKaurov)

@Szabolcs Not for me. What was it, do you remember?

@Szabolcs Thanks for reporting. We have some corner cases after recent update. Content is there but does not show. This should be fix sometime soon.

@VitaliyKaurov I think it was a fast C++ / LibraryLink implementation of BinCounts, as well as some tips on how to get a faster BinCounts with plain Mathematica code. As I remember, I started the thread, and Sander posted several replies too.

There are multiple such empty threads now.

Should I tell you if I see more?

There is a lot of valuable content on W|C. I hope it will not get lost.

Maybe it is not a good thing that W|C is fully managed by WRI.

The old MathGroup was never perceived as independent, but at least it was a Usenet group. That means that the contents are archived forever, and also independently.

The W|C forum backend is so wonky, I am worried that things will go really wrong eventually.

I should check the Internet Archive.

@Szabolcs do not worry, the Dbs are backed up, all info is there actually, just front end glitch. do you remember what date that thread was started?

roughly

web.archive.org/web/20140420075011/https://…

The snapshot doesn't have the LibraryLink stuff, but it does have a pure-Mathematica fast BinCounts.

4 years ago about

@Vitaliy 2014 April 16 is my best guess based on the above (4 days old, snapshot time Apr 20)

yeah those are corner cases

this will be fixed soon

OK. If I see more of these, should I send them to you? Or will they get sorted out eventually without feedback?

yes pllease ! sorry for inconvenience

yes they all will resurrect automatically

but good to have links for testing

Is there a way to do a regression using a neural network with variable input size? I can control the size of input while training but I would have a hard time trying to controlling input size while testing or in real time use.

I am just trying to use bunch fully connected layers with ReLU or tanh as activations.

@Szabolcs very custom xml and ready enough that built-in xml->html works well for it.

@GalAster Thanks