Continuous Normalizing Flow
A generative model that defines a continuous transformation from a prior distribution to a data distribution via ordinary differential equations, offering invertibility and exact likelihood computation.
Core Facts
Timeline (last 90 days)
Common normalizing flow architectures include Real-NVP, Neural Autoregressive Flow (NAF), and Continuous Normalizing Flow (CNF), each with trade-offs in computational efficiency and expressiveness.
The idea of flow matching is to learn a time-dependent vector field v_t(x) such that the ordinary differential equation driven by this vector field can transport a simple reference distribution to the target data distribution.
All Facts (2)
The idea of flow matching is to learn a time-dependent vector field v_t(x) such that the ordinary differential equation driven by this vector field can transport a simple reference distribution to the target data distribution.
75%UnverifiedCommon normalizing flow architectures include Real-NVP, Neural Autoregressive Flow (NAF), and Continuous Normalizing Flow (CNF), each with trade-offs in computational efficiency and expressiveness.
50%