Structured Gaussian Variational Inference
Descriptions may be stale. Consult the linked Lean sources.
Structured Gaussian Variational Inference
We optimize Gaussian location
Let
Our assumptions:
Assumption 1 (Smoothness and strong monotonicity).
The potential
Assumption 2 (Attainment and initial distance).
There exists
Queries may be made at any
Results
On
| Result | Oracle complexity | Novelty | Limitations |
|---|---|---|---|
| (1) StructuredGaussianVI
(Ours) |
|||
| (2) (Kim et al. 2025), Thm. 3.2, Cor. 3.3, Prop. 3.4; mean-field, Gaussian base | (1): worst-case
statistical dependence |
(1): polynomial |
|
| (3) (Kim et al.
2025), Thm. 3.2, Prop. 3.4; mean-field Gaussian, constant |
(1): for |
(1): can lose |
|
| (4) (Ko et al.
2025), Thm. 3; coupled target, |
(1): worst-case
statistical saving about |
Abbr:
| Result | Oracle complexity | Novelty | Limitations |
|---|---|---|---|
| (5) (Ko et al.
2025), Thm. 3; decomposed |
(1): averaged KL guarantee. No established complexity improvement on decomposable targets; different primitive oracles. | (1): may lose when |
|
| (6) Repository full-rank
result (ReparameterizedGaussianVI
2026); |
(1): scope extends to
every sparse Cholesky family, with its own minimizer. Matches known rate
at |
Abbr: