Hyperparameter Learning for Latent Factorization of Tensors for Representation Learning to Large-scale Dynamic Weighted Directed Network
Researchers have developed an automated framework using Differential Evolution (DE) to optimize hyperparameters for Latent Factorization of Tensors (LFT) models. This DE-LFT method aims to reduce the manual effort and computational resources typically required for tuning LFT models, which are used to analyze large dynamic networks. By integrating DE into the LFT training process, the framework adaptively searches for optimal regularization parameters, leading to improved prediction accuracy as demonstrated on real-world datasets. AI
IMPACT Automates a key step in applying complex models to dynamic network data, potentially reducing barriers to entry for researchers.