Scalable Sparse Autoregressive Variational Inference for Non-Gaussian Spatial Fields
We present a scalable variational inference framework that mirrors the sparse nearest-neighbor autoregressive structure of the prior in the variational family, uses normalizing-flow conditionals for expressiveness, and amortizes all variational parameters via a single shared neural conditioner.
ProofPaper ↗
Key points
- Latent field models are essential for spatial statistics, but the posterior distribution of the latent field, even with a Gaussian process (GP) prior, is analytically intractable whenever the likelihood is non-Gaussian.
- Existing approximations are either too slow or too restrictive for large non-Gaussian datasets.
- Doubly stochastic optimization with reduced ancestor sets achieves polylogarithmic complexity per iteration.
Sources (1)
- [1]Scalable Sparse Autoregressive Variational Inference for Non-Gaussian Spatial FieldsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 02:30 PM
We present a scalable variational inference framework that mirrors the sparse nearest-neighbor autoregressive structure of the prior in the variational family, uses normalizing-flow conditionals for expressiveness, and amortizes all variational parameters via a single shared neural conditioner.
Latent field models are essential for spatial statistics, but the posterior distribution of the latent field, even with a Gaussian process (GP) prior, is analytically intractable whenever the likelihood is non-Gaussian.
Extractive summary: sentences quoted from the sources.