An explainer on Tree Ensemble Layer

Both Neural Networks and Decision Trees have worked exceptionally on multiple machine learning problems. What if we could get the best of both in one model? This is what the Google Research team tried to seek in their paper titled "The Tree Ensemble Layer: Differentiability meets Conditional Computation". Trees support conditional computation i.e. they are able to route each sample through a small number of nodes. This can lead to both performance benefits, enhanced statistical properties, and helps in interpretability.