deddy

4 ms·

1.
by deddy 10mo ago
There isn’t really much benefit to having compute on orbit unless you’re working on VERY specific applications that have such tight latency requirements where you need to process the data immediately as it comes out of the sensor. In which
2.
by deddy 10mo ago
It depends on what you're going for. Generally people really don't want collisions due to cascading effects, so they take the worst-case probability of collision found with bounding assumptions. Additionally, while often all these
3.
by deddy 10mo ago
This is confusing terminology in the field, but you generally talk about the cross sectional area in the plane of the conjunction ( https://www.space-track.org/documents/SFS_Handbook_For_Opera... ) to calculate the proba
4.
by deddy 10mo ago
Need to do a full read in more depth but it looks like they used a collision cross section of A=300 m^2, which is a little conservative but not insane given that the current Starlink v2 mini has about 90-120 m^2 of total surface area on its
5.
by deddy 2y ago
So other ways of combining greater numbers of microbatch gradients in an effective/consistent manner for performing an update is one area of potential future work. I think your idea is an interesting way to approach it. Though there ar
6.
by deddy 2y ago
> 1. Do you think this is a form of variance reduction or more a form of curriculum (focus first on the bulk, then on remaining errors)? I'd say generally more of a cirriculum (using your terminology). Broadly speaking the idea is t
7.
by deddy 2y ago
Great question! There are a few different aspects to this. With gradient agreement filtering having a greater number of batches (generally) increases the likelihood of finding another microbatch that agrees with the gradient simply by virtu
8.
by deddy 2y ago
Co-author here. Super excited to see this work posted on HN! Happy to answer questions.
9.
by deddy 7y ago
F10.7 is the measured solar power flux at 10.7 cm wavelength. It is measured all over the world but primarily from Penticton observatory in Canada. The largest source of uncertainty in orbit propagation for satellites in Low-Earth Orbit (LE
10.
by deddy 7y ago
Currently VC backed companies are barred from competing on SBIRs. There is legislation currently in the works to change that. But gov funding isn’t more common currently because you legally can’t do it. Edit: I need to amend that to say maj