Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Deep Heteroskedastic Regression: Post-Hoc Variance Estimation from Latent Representations

NeurIPS 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Single-model predictive uncertainty

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes post-hoc variance estimation for deep heteroskedastic regression by fitting a variance model on latent representations of a trained network, avoiding joint mean-variance training.

Problem

Training deep neural networks to jointly parametrise mean and variance for heteroskedastic regression is difficult, motivating uncertainty quantification that is more reliable and easier to train.

Approach

Estimates predictive variance post hoc from the latent representations of a network already trained for the mean, rather than training mean and variance heads jointly end to end.

Closest papers by meaning

• Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.88)
• Towards Self-Supervised Covariance Estimation in Deep Heteroscedastic Regression (ICLR 2025 · similarity 0.87)
• Uncertainty Quantification for Deep Regression using Contextualised Normalizing Flows (NeurIPS 2025 · similarity 0.87)
• TIC-TAC: A Framework For Improved Covariance Estimation In Deep Heteroscedastic Regression (ICML 2024 · similarity 0.86)
• Fully Heteroscedastic Count Regression with Deep Double Poisson Networks (ICML 2025 · similarity 0.86)
• Uncertainty Quantification for Forward and Inverse Problems of PDEs via Latent Global Evolution (AAAI 2024 · similarity 0.86)