Starbeamrainbowlabs
LightNews LightNews
banner
sbrl.fediscience.org.ap.brid.gy
Starbeamrainbowlabs
@sbrl.fediscience.org.ap.brid.gy
7 followers 0 following 26 posts
Computer Science PhD Student Researcher @ University of Hull, I love to learn 🌟

#OpenSource • #Linux • #Sysadmin • #AI • #Piano • #Parkour • #Cats […]

[bridged from https://fediscience.org/@sbrl on the fediverse by https://fed.brid.gy/ ]
Posts Replies Media Videos
sbrl.fediscience.org.ap.brid.gy
Fixed it 💖

https://gitlab.com/sbrl/bin/-/raw/master/tinyplanet

Blog post coming soon..... just as soon as I have the energy to write it :D

#photography #photoedit #hyperfocus #specialInterest #bash #linux #binfolder
A photograph of some green hills and fields in the countryside below a cloudy sky, but turned into a tiny planet. Because the source photo was mostly sky, the planet's radius is only ~1/8th of the width of the image, leaving the clouds above to swirl around it. The edges of the images copy the outermost pixel outwards, creating subtle streaks outwards to frame the image.
May 27, 2025 at 10:15 PM
sbrl.fediscience.org.ap.brid.gy
While I try to find time to write a PhD update blog post, here's a Cool Graph I plotted recently for my rainfall radar model.

One of the things I have been asked to do in my PhD corrections is to assess the model's stability. To this end, I trained a series […]

[Original post on fediscience.org]
A Cool Graph showing validation accuracy over training for my rainfall radar model. Unsurprisingly, a lower learning rate (lr=0.00001) results in higher and more stable validation accuracy than a higher learning rate (lr=0.01). Particularly higher learning rates are very variable in accuracy; stretching from 0.3 - 0.8. Lower learning rates are ~0.8 - ~0.95.
March 3, 2025 at 1:37 PM