Deep Forests

Deep Forests can be viewed as computational pipelines that act differently depending on execution stage: fit or transform. In contrast to neural networks such pipelines don’t use backpropagation at the training step.

For managing such structures we’ve devised a new framework called bosk. It can be used to build classical Deep Forests as well as arbitrary machine learning model pipelines.