Normalizing flows

训练方式

Maximum likelihood is optimized exactly under architectural constraints that make inversion and the Jacobian determinant tractable.

推理运行方式

Density evaluation runs data toward the latent; generation samples the base distribution and applies every transform in reverse.

优势

• •Exact likelihood under the model
• •Invertible encoding and generation
• •Useful when density estimation is itself important

权衡

• •Invertibility constrains network design
• •High-dimensional media can require deep, memory-heavy flows
• •Likelihood does not necessarily track perceived sample quality

适用场景

• Exact density or reversible transforms are requirements
• The domain fits available invertible architectures
• You will evaluate both likelihood and task utility

应避免或质疑的场景

• Only perceptual generation quality matters
• Architectural flexibility is more important than exact likelihood
• A simpler discriminative uncertainty method is sufficient