Extended Data Fig. 2: Combining latent features of separately trained models. | Nature Machine Intelligence

Extended Data Fig. 2: Combining latent features of separately trained models.

From: Simultaneous dimensionality reduction and integration for single-cell ATAC-seq data using deep learning

Extended Data Fig. 2

Three ensembles consisting of ten VAE models were fitted on the Buenrostro et al. 2018 dataset. The total loss across the dataset was determined for each model and models with poor outlier losses were excluded from the ensemble (for example, due to poor local minima; see Methods). Individual models (BAVARIA - individual) were combined to ensembles by concatenating the latent features (BAVARIA - ensemble). Latent features were subjected to clustering using several algorithms and clustering performances were computed based on adjusted mutual information (AMI), adjusted Rand index (ARI) and Homogeneity (Hom). The best score across the clustering algorithms are considered.

Back to article page