Fisher Information Metric
A Riemannian metric on statistical manifolds that measures how much a probability distribution changes in response to small perturbations of model parameters. It is coordinate-invariant and can be applied in optimization contexts such as neural network pruning.
Timeline (last 90 days)
Bures geometry is a natural Riemannian metric defined on the space of quantum states (density matrices), induced by the Bures distance; on classical probability distributions it is closely related to the Fisher information metric; in the quantum case, it measures the minimum statistical distinguishability between two quantum states.
The paper proposes using geodesic distance determined by the Fisher information metric as a measure of the true change brought about by pruning.
The Fisher information metric is a Riemannian metric on the statistical manifold, proposed by statistician Ronald Fisher, and has the property of being coordinate-independent.
All Facts (3)
Bures geometry is a natural Riemannian metric defined on the space of quantum states (density matrices), induced by the Bures distance; on classical probability distributions it is closely related to the Fisher information metric; in the quantum case, it measures the minimum statistical distinguishability between two quantum states.
50%UnverifiedThe paper proposes using geodesic distance determined by the Fisher information metric as a measure of the true change brought about by pruning.
50%UnverifiedThe Fisher information metric is a Riemannian metric on the statistical manifold, proposed by statistician Ronald Fisher, and has the property of being coordinate-independent.
50%