Khushi

@ChinmayKak Thank you❤️

Congratulations @starlitmatcha IEEE first author it is🥳🥳🥳🥳♥️♥️♥️

New blog post ! This one is related to my recent experiments on slowrun with various papers as experiments, and a deep dive on hybrid attention and it's importance for slowrun, and in general data efficiency. Please do check it out, has some cool visuals;) RT/QT for the algo!

Midsems over🥹

What a week

New blog post! This is related to the pretty cool slowrun repository by @industriaalist and @akshayvegesna, what i tried, what worked and what failed! More of a dev log of my experiments than a super polished work, but please do check it out anyways. RTs/QTs always welcome:)

Blog link: @TVSokkwlRzSyZ22QjIUs4w/HJ1QPUbOWx" target="_blank" rel="nofollow noopener">hackmd.io/@TVSokkwlRzSyZ…

Hello everyone! I have been recently exploring RL and did some experiments and visualizations to help grow my intuition about it, and have written this blog as a record of the same. I have also linked the code and blog link below Please do check it out and RTs/QTs are welcome:)

Because of that, rotation and translation can be done together with one matrix multiplication. So instead of rotate → move → rotate → move, you just multiply matrices and see where the robot ends up. Cool how you simplify things.

Was studying homogeneous transformation matrices in Robot Kinematics today. In robotics every robot link has its own coordinate frame, so it becomes difficult to figure out how motion looks from one frame to another.

This is exactly how out like a light by the honeysticks feels like

this was the first pcb i designed in first year

making ant do crab rave

Do check it out!

@ChinmayKak ♥️♥️♥️

Happy Valentines day @starlitmatcha ♥️♥️(painted by her)

@ChinmayKak Lesssgooo

First of many 🤞

Throwback to last year when we were trying to get our robot to make a dunk shot for Robocon '25

Next will implement these learnings.

Spent time with Q-learning and value-based RL today. Fun thing I noticed: the Bellman equation follows the same logic as dynamic programming.

Also understood the contrast: Monte Carlo waits until the episode ends whereas TD learning updates the value function step by step. Q-learning works because it uses these TD updates. So you don’t pick actions directly. You learn Q-values and choose actions based on them.