Hacker News new | ask | show | jobs
Swap Curve Fitting (colab.research.google.com)
32 points by thrower 1923 days ago
4 comments

For some additional background. This is a demo from the quant finance tensorflow package:

https://github.com/google/tf-quant-finance

Performance seems impressive relative to other FOSS. I use quantlib for mostly for hobby projects, would love to hear peoples thoughts if they've played with this.

I thought this was a Google paper since the domain shows up as "research.google.com", but it's actually a random Colab notebook.
It is a Google paper, further along in the URL one can see that it's under the offical Google github account.
It's a demo of a Google library:

https://github.com/google/tf-quant-finance

as a former rates quant, this is cute
Why? Is this unusable naive or why would they build it? Excuse me for my lack of knowledge, far from my field.
As far as i can tell, their swap curve construction doesn't support dual curve discounting [1], which makes this a toy.

There's no reason someone couldn't make this library industrial-strength, but is anyone going to? Is this supposed to be a supported product, or is this basically a tech demo to show what TensorFlow is capable of?

[1] https://www.crd.com/insights-valuing-interest-rate-swaps-the...

What are you trying to say?
I find your commment much harder to understand than the post. Are you an experienced quant who finds the work trivial, or an undergrad with no idea what a swap curve is? It's impossible to tell and thus impossible to answer your question.
Neither, this reads like a post to signal one's knowledge rather than one's intelligence. Do you want so see that finite state transducer I wrote in C?