f(θ) = ½ θᵀHθ
θ⋆
κ(H) ≫ 1
gradient descent
momentum
∇θ L(θ)
x ∈ ℝⁿ
hθ
Mθ = { hθ(x) : x ∈ X }
∂hθ / ∂z₁
∂hθ / ∂z₂
hθ(x₀)
Home
Work
Blog
Note
Research
Projects
Notes
biology
01
A prologue to protein structure prediction
Aug 8, 2026
Stable
Nearly slop
reinforcement learning
01
Note on Sutton & Barto — Chapter 3: Finite Markov Decision Process (MDP)
Jul 22, 2026
Stable
Nearly slop
mathematics
01
A Little Bit About "Numbers" and "Mathematics"
Jul 19, 2026
Pending
Nearly slop
llm
01
Residual Stream is Key to Transformer Interpretability
Jul 18, 2026
Stable
Nearly slop