Why it matters
Shampoo is second-order optimizer for large models. Understanding shapes advanced training.
Advertisement
The architecture
Per-layer preconditioners (L, R matrices).
Root of preconditioners.
Update with preconditioned gradient.
Advertisement
How it works end to end
Gupta et al. 2018.
Preconditioner: outer product of gradients.
Fractional matrix power.
Distributed Shampoo scales.