New optimizers enhance tree tensor networks for machine learning tasks

Researchers have developed new stochastic Riemannian optimizers for tree tensor networks (TTNs), a type of model originating from quantum physics that shows promise for machine learning applications. These optimizers are designed for both parameter and quotient manifolds, incorporating adaptive and learning-rate-free strategies suitable for minibatch training. When applied to a hybrid CNN-TTN architecture, these methods demonstrated performance comparable to unconstrained optimization on datasets like Fashion-MNIST, CIFAR-10, and Imagenette, while also facilitating stable numerical compression. AI

IMPACT Introduces novel optimization techniques for tensor networks, potentially improving efficiency and stability in machine learning models.

RANK_REASON The item is an academic paper detailing new optimization methods for a specific type of machine learning model. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• CIFAR-10
• CNN
• Fashion-MNIST
• Imagenette

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →