WRSWorld Research StateContribute
WORLD RESEARCH STATE

NICE, Real NVP and Glow: coupling-based generative flows

Contribution grade

Flow-based probabilistic generative modeling · Human-designed research method or resource

Reference score200,000Provisional editorial estimate of this exact contribution; its grade is assessed separately.

What changed

Developed tractable invertible generative models through additive coupling, affine coupling with learned scale, and Glow's invertible1x1 channel mixing and ActNorm within multiscale architectures.

Scientific scope

Developed tractable invertible generative models through additive coupling, affine coupling with learned scale, and Glow's invertible1x1 channel mixing and ActNorm within multiscale architectures. A chronological family, not one joint paper. Exact density is under the continuous model, not proof of matching real data; image dequantization/preprocessing matters. NICE's unit-determinant additive blocks, Real NVP's affine blocks and Glow's channel mixing remain separately attributed. Invertible1x1 weights must remain nonsingular; ActNorm is data-initialized, not ordinary batch normalization.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. A chronological family, not one joint paper. Exact density is under the continuous model, not proof of matching real data; image dequantization/preprocessing matters. NICE's unit-determinant additive blocks, Real NVP's affine blocks and Glow's channel mixing remain separately attributed. Invertible1x1 weights must remain nonsingular; ActNorm is data-initialized, not ordinary batch normalization.

Contributors

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction