GammaNet - Stable Feature-Space Decay in Linear RNNs
GammaNet - Allowing Kimi Delta Attention to decay along meaningful directions instead of meaningless coordinates.
Lucas Sun
I write about math, machine learning, and whatever else I find interesting.
Bra-Ket (Dirac) Notation in Deep Learning (Part 1)
Bra-ket notation as a unifying language for linear maps, MLPs, and self-attention — making the key-value structure of deep learning explicit.
First Post
This is my first post. I put off writing this for too long.
Hello World
Hi! I'm Lucas Sun. Welcome to my blog.