Alternating Local Enumeration (TnALE): Solving Tensor Network Structure Search with Fewer Evaluations
Li, Chao, Zeng, Junhua, Li, Chunmei, Caiafa, Cesar, Zhao, Qibin
–arXiv.org Artificial Intelligence
Tensor network (TN) is a powerful framework in machine learning, but selecting a good TN model, known as TN structure search (TN-SS), is a challenging and computationally intensive task. The recent approach TNLS~\cite{li2022permutation} showed promising results for this task, however, its computational efficiency is still unaffordable, requiring too many evaluations of the objective function. We propose TnALE, a new algorithm that updates each structure-related variable alternately by local enumeration, \emph{greatly} reducing the number of evaluations compared to TNLS. We theoretically investigate the descent steps for TNLS and TnALE, proving that both algorithms can achieve linear convergence up to a constant if a sufficient reduction of the objective is \emph{reached} in each neighborhood. We also compare the evaluation efficiency of TNLS and TnALE, revealing that $\Omega(2^N)$ evaluations are typically required in TNLS for \emph{reaching} the objective reduction in the neighborhood, while ideally $O(N^2R)$ evaluations are sufficient in TnALE, where $N$ denotes the tensor order and $R$ reflects the \emph{``low-rankness''} of the neighborhood. Experimental results verify that TnALE can find practically good TN-ranks and permutations with vastly fewer evaluations than the state-of-the-art algorithms.
arXiv.org Artificial Intelligence
May-29-2023
- Country:
- South America > Argentina (0.04)
- North America > United States
- Hawaii > Honolulu County > Honolulu (0.04)
- Asia
- Middle East > Republic of Türkiye
- Karaman Province > Karaman (0.04)
- Japan > Honshū
- Kantō > Tokyo Metropolis Prefecture > Tokyo (0.14)
- China
- Heilongjiang Province > Harbin (0.04)
- Guangdong Province > Guangzhou (0.04)
- Middle East > Republic of Türkiye
- Genre:
- Research Report (0.63)
- Technology: