Publications by authors named "Matt Kaeberlin"
Article Synopsis
- There's a rising interest in using unsupervised deep learning for gene expression analysis, leading to the development of methods to improve model interpretability.
- These interpretability methods fall into two categories: post hoc analyses of complex models and the design of biologically-constrained models from the start.
- The authors suggest that combining these two approaches can be beneficial and introduce PAUSE, a method that pinpoints key sources of transcriptomic variation using both unsupervised learning and biologically-constrained neural networks.
