Discovering Nonlinear Static Relationships in Unlabeled Dataset using Autoencoder with Ordered Variance
Researchers have developed an autoencoder with ordered variance (AEO) that enhances the latent space structure by ordering latent variables based on their variance. This method allows for the systematic determination of latent space dimensionality and the discovery of nonlinear relationships within unlabeled datasets. The framework, extended to a ResNet-based version (RAEO), enables unsupervised static model extraction and has been demonstrated on a continuous stirred tank reactor process for real-time optimization. AI
IMPACT Introduces a novel unsupervised method for uncovering complex relationships in data, potentially improving model interpretability and extraction.