Backstaff Parameter Tuner banner

Backstaff Parameter Tuner

6 devlogs
48h 12m 24s

This is a novel algorithm that bypasses backpropogation and makes it super easy to tune parameters that you cannot necessarily take the gradients of - think PID loops in a robot or ingredients in a recipe. It is all wrapped up in a nice little UI …

This is a novel algorithm that bypasses backpropogation and makes it super easy to tune parameters that you cannot necessarily take the gradients of - think PID loops in a robot or ingredients in a recipe. It is all wrapped up in a nice little UI so you know what is going on too!

This project uses AI

I occasionally used Copilot to help with some funky QT stuff, but the whole algorithm is written by me.

Demo Repository

Loading README...

Max Clemetsen

Shipped this project!

I built an app that tunes parameters when you tell it how well it performs! It was tricky to actually get good gradient estimates, but it worked out in the end and I think it could be pretty revolutionary.

Max Clemetsen

TESTING ALL OF THIS STUFF ON AN ACTUAL CLANKER LETS GOOO!!! (the ui works but the algo is actually kinda scuffed so i have to fix that ig)

Attachment
0
Max Clemetsen

Adding some cool vision processing stuff! Got some prototyping and testing done too.

Attachment
0
Max Clemetsen

Started working on some CAD for the robot and added some work for the actual tui,

Attachment
0
Max Clemetsen

Initial devlog. Nothing special here. Basic algo is in place, but it is far from good! Idk why this devlog includes time from before the challenge started tho lol

0