Stochastic Optimization of Tree Tensor Networks

Name: Stochastic Optimization of Tree Tensor Networks
Uploaded: 2026-10-06T10:51:50.084544+00:00
Duration: 309 s
Description: The paper develops stochastic Riemannian optimizers for Tree Tensor Networks (TTNs) by optimizing them directly on their geometric manifolds rather than in standard Euclidean space. It introduces methods like RADAM and RDOG to handle the parameter and quotient manifolds, achieving numerically stable training and better adherence to the TTN's hierarchical structure during optimization.

The paper develops stochastic Riemannian optimizers for Tree Tensor Networks (TTNs) by optimizing them directly on their geometric manifolds rather than in standard Euclidean space. It introduces methods like RADAM and RDOG to handle the parameter and quotient manifolds, achieving numerically stable training and better adherence to the TTN's hierarchical structure during optimization.

Open the full episode