Paper表征形成的信息几何
The Information Geometry of Representation Formation
A research paper that proposes analyzing the process of neural network representation formation using Bures geometry and quantum Fisher information, treating representation learning as a geometric trajectory and revealing relationships among sensitivity concentration, trajectory curvature, and computational complexity.
Timeline (last 90 days)
Oct 27
A study titled "The Information Geometry of Representation Formation" proposes viewing representation learning as geometric trajectories, using Bures geometry to measure movement in representation space and quantum Fisher information to characterize local sensitivity to input factors.
Unverified50%