See for instance the pytorch geometric [1] package, which is the main implementation in pytorch. They also link to some papers there that might explain you more.
I agree, combining GNNs with temporal signals and dynamics graphs make them so expressive.
Another great recent paper on this topic https://arxiv.org/abs/2310.15978
> If you want to predict (multiple) time series using multiple series as input/predictors, that's a whole new level of difficulty. I don't know of a good automatic/fast/scalable approach that properly guards against overfitting
We only did a few internal NN and LSTM implementations in the past, we should probably evaluate the new pytorch stuff soon. But as you can imagine a lot of our time was consumed by modelling pandemic-induced dynamics (which is especially at longer forecast horizons are much more driven by assumptions rather than by data/models).
I think the OP is referring to algorithmic probability. See
[1] R. J. Solomonoff. A formal theory of inductive inference: Parts 1 and 2. Information and Control, 7:1--22 and 224--254, 1964.
[2] R. J. Solomonoff. Complexity-based induction systems: Comparisons and convergence theorems. IEEE Transactions on Information Theory, IT-24:422-432, 1978.
I'm not making any claims about whether the notation is good or bad, but just because something is the standard doesn't mean it's therefore good.
And the person you're replying to seems to be well aware it is a standard notation: "Can I rant on how bad much mathematical notation is in terms of usability?"
Open source really is irrelevant. We don't know what code they deployed - it could be an open source project (we hope so), it might be an open source project with some minor changes - an NSA-approved backdoor for example, etc. We don't have any guarantee, so far as I'm aware, of what they're running, much less that it does what it claims to do.
Open source makes it is easier for security experts to review the code and determine whether it meets its security claims. If it is closed source, then there is no guarantee that the code the experts review is the same code that is used by the service. And also it is at the company's discretion whether to allow a security audit or not, and then which auditors to allow or exclude.
But if the whole of the client software is open source, then it also eliminates the need to trust binaries provided by the vendor (which can be MITM) because you can build the software yourself or use a version audited by an entity that you do trust.
That, also the fact that if they have it in their privacy policy and website, then we can call them out on it when they insert a backdoor or remove the encryption. We can't say the same about Whatsapp and its end-to-end encryption, because they never even publicly admitted to using it. How can we ever hold Whatsapp responsible for not using end-to-end encryption then?
