Higher-order organization of multivariate time series | Hacker News Reader

Higher-order organization of multivariate time series

arxiv.org

80 pointsAnon8415 comments

Old explainer thread on twitter from one of the authors: https://twitter.com/lordgrilo/status/1506294750621716482

He mentions a lot of recent similar efforts. One thing I don’t see is Martin Zwick’s continuation of reconstructability analysis / discrete multivariate modeling.

e.g. https://pdxscholar.library.pdx.edu/sysc_fac/22/

> RA was derived from Ashby (1964), and was developed by Broekstra, Cavallo, Cellier Conant, Jones, Klir, Krippendorff, and others (Klir, 1986, 1996).

(Papers are pretty nowadays)

Scalable High-Order Gaussian Process Regression

https://proceedings.mlr.press/v89/zhe19a.html

How is this related?

I'm not sure, my intuition might be wrong.

I looked into this paper and expected Gaussian processes with complex kernel functions.

Giovanni Petri (author of paper here):

https://twitter.com/lordgrilo/status/1506294750621716482 ->

Networks beyond pairwise interactions: Structure and dynamics https://reader.elsevier.com/reader/sd/pii/S0370157320302489?... ->

Multiscale Information Decomposition: Exact Computation for Multivariate Gaussian Processes https://www.mdpi.com/1099-4300/19/8/408

So I guess if telepathy is possible, there should be a Higher Order indication for people who share a link huh?

And yes, in finance, the correlations between asset classes shoot up toward 1 in periods of crisis (black swan event) . Hence, the research for tail-hedging strategies...

> And yes, in finance, the correlations between asset classes shoot up toward 1 in periods of crisis (black swan event) . Hence, the research for tail-hedging strategies...

Related to what you said here, I was surprised there wasn't a comparison with Vine Copulas in the paper or thread! But this is pretty far outside of my realm of expertise, so maybe it shouldn't be surprising.

It took us a long time to figure out 0. This idea of multivariate time series is "tacit knowledge" and once we get good at modeling it we will never look back. As a laymen this is the most accessible I've seen these concepts.

And speaking for neurology methodology for time series tests subtraction sucks! https://www.researchgate.net/publication/12369885_How_to_Avo...

Can you say more? What's "0" in this context?* and can you explain what the idea here is and why it's useful? I think quite a few readers would be curious.

(* I ask because I can't find anthing labeled 0 - but I may have messed up the URL - see https://news.ycombinator.com/item?id=34223587)

I mean zero as in the nothing sense. We had trade and society before we had way to mathematically describe nothing. The history is diverse and interesting:

https://en.m.wikipedia.org/wiki/0

before we had [a] way to mathematically describe nothing.

This seems a little overstated since positional number systems are fairly sophisticated mathematics - you can deal with 'nothing' without needing zero in its modern sense.

You can even have positional number systems without zero - that's what the Babylonians had

Unveiling the higher-order organization of multivariate time series:

https://arxiv.org/abs/2203.10702

Ok, since https://www.nature.com/articles/s41567-022-01852-0 is hardwalled and that arxiv.org post has the same authors and almost the same title, I'm going to call it close enough and replace the URL above. Thanks!

If there's a better URL that people can openly read, we can change it again.

This link will allow people to read it in the browser without subscription:

https://rdcu.be/c2DJE

Edit: Please let me know if it doesn't work

Reply on news.ycombinator.com