Kähler Geometry of Quiver Varieties and Machine Learning
arXiv:2101.11487
Abstract
We develop an algebro-geometric formulation for neural networks in machine learning using the moduli space of framed quiver representations. We find natural Hermitian metrics on the universal bundles over the moduli which are compatible with the GIT quotient construction by the general linear group, and show that their Ricci curvatures give a Kähler metric on the moduli. Moreover, we use toric moment maps to construct activation functions, and prove the universal approximation theorem for the multi-variable activation function constructed from the complex projective space.
arXiv commentv2: expanded on related works. 47 pages, 13 figures
References in corpus (12)
- Ricci flow with surgery on three-manifolds
- Group Equivariant Convolutional Networks
- Identifying and attacking the saddle point problem in high-dimensional non-convex optimization
- Homogeneous coordinate rings and mirror symmetry for toric varieties
- Symplectic quotients by a nonabelian group and by its maximal torus
- On the saddle point problem for non-convex optimization
- Learning Activation Functions: A new paradigm for understanding Neural Networks
- Natural Graph Networks
- Immersed two-spheres and SYZ with Application to Grassmannians
- Covariance in Physics and Convolutional Neural Networks
- Graph Laplacians, Riemannian Manifolds and their Machine-Learning
- The Representation Theory of Neural Networks