Stein Operator
A differential operator acting on functions, used in the Stein method to construct valid statistics that do not depend on normalization constants; its core component is the score function.
Timeline (last 90 days)
This research models the log density ratio between the target distribution and the base distribution as the negative of the base distribution's Stein operator applied to the displacement field, plus a normalization constant.
Stein Variational Gradient Descent (SVGD) takes a set of particles starting from an initial distribution and iteratively moves them along a velocity field derived from the Stein operator, ultimately making the particle distribution approach the target distribution; its update direction is the gradient of the kernelized Stein discrepancy (KSD), which is essentially a displacement field.
The effectiveness of this method depends on the choice of the base distribution and the specific form of the Stein operator, and it may face challenges in scenarios with large distribution shifts.
All Facts (3)
This research models the log density ratio between the target distribution and the base distribution as the negative of the base distribution's Stein operator applied to the displacement field, plus a normalization constant.
50%UnverifiedStein Variational Gradient Descent (SVGD) takes a set of particles starting from an initial distribution and iteratively moves them along a velocity field derived from the Stein operator, ultimately making the particle distribution approach the target distribution; its update direction is the gradient of the kernelized Stein discrepancy (KSD), which is essentially a displacement field.
50%UnverifiedThe effectiveness of this method depends on the choice of the base distribution and the specific form of the Stein operator, and it may face challenges in scenarios with large distribution shifts.
50%