Pseudomanifold

Darktable (https://www.darktable.org/) has become very good and continues to be my go-to solution for digital photography.

Hi HN,

We are very excited to share a new dataset chock full of interesting triangulations with you. In machine learning, a lot of works try to handle such higher-order inputs, but we show that there is still a long way to go. Let us know what you think!

Great selection of works, but I am missing a lot of references from topology in ML, with the article only assuming a very cursory perspective in terms of 'topology captures connectivity and/or continuity.'

Some works from my colleagues and me go a little bit deeper (no pun intended), for instance:

- Neural Persistence Dynamics: https://arxiv.org/abs/2405.15732

- Simplicial Representation Learning with Neural $k$-Forms: https://openreview.net/forum?id=Djw0XhjHZb

- A general review on topology in machine learning: https://www.frontiersin.org/journals/artificial-intelligence...

There are more things in topology and machine learning, Horatio, than are dreamt of in your article ;-)

My condolences for your loss. May his memory be a blessing.

Author here: Fixed, thanks a lot!

Maybe 'assumption' would have been the better choice. Given that HN is (for me!) broadly about playfully engaging with technology and society at large, I would expect that this _should_ attract a relatively diverse audience.

Fully agree with you there! As I said, I'm merely interested in getting a rough picture of the demographics (INB4: selection bias etc.). I find the statement you mention just as problematic as if you were to switch out the ethnic identity with _any_ form (self-identified) identity. That being said, I don't see anything problematic about looking at population demographics and then asking _why_ some identities are over- or under-represented. This poll should not be used for that, though :-)

I agree that this will be biased and I don't think this should be used for any serious type of survey. I was literally just interested in the results. :-)

Thanks, will do! It is not relevant for the discussions as such, but still an interesting aspect of HN to study, IMO.

Pleasant in the sense that I'd love to see some myths about this being a 'man-only' site dispelled. Should have clarified that!

They missed a great opportunity to call this BloTorch (Bayesian Learning and Optimisation?) here, but I'm very excited to see such methods gaining more traction!

I agree with this sorry state of affairs; it does not seem that there is a proper successor to X at the moment.

Thanks for everything, Bram. Your memory will be a blessing.

Amazing to see 'magnitude' on the front page of Hacker News! If you are interested in a direct application of this invariant beyond ecology, check out our recent pre-print in which we study the generalisation behaviour of neural networks: https://arxiv.org/abs/2305.05611

My personal approach to magnitude is not based on category theory but rather based on weightings of a metric space. If your metric satisfies certain properties, you can obtain a measure of the 'effective number of points' of a metric space. This is particularly relevant when looking at the metric space from different scales---zooming in gives you a lot of disconnected points, while zooming out gives you clusters. Magnitude then captures the changes in the number of points in a principled manner.

Here's mine:

https://bastian.rieck.me/blog/

I mostly write about academia and machine learning these days, but every once in a while, I also have the urge to write a really nerdy post on a more technical topic. Writing continues to be cathartic for me, and I hope to make a small difference when I discuss things that are not typically discussed openly (in an academic setting).

Feedback is very welcome!

Author here :-) Thanks a lot---the font is Gentium Book Basic. If you are interested in the layout, send me an e-mail or a DM; I can share the theme.

I use this on several of my sites as well. None of them get any significant amount of traffic, so I figured it would be okay to add more names to the list, as a small token of keeping their memory alive.

Browsing the web with the Clacks Overhead extension for Chrome (https://chrome.google.com/webstore/detail/clacks-overhead-gn...), I'm always surprised by which sites enable this extension. My latest find is arXiv.org! If you go to any abstract page, such as https://arxiv.org/abs/hep-th/0310077v2 (picked that one at random for its humour), the header will be sent.

Wow, did not know that this existed. Thanks so much!