Is deep learning a stream?
Deep learning is defined as a (potentially infinite) set of network topologies with interactions between nodes in these networks governed by math functions, input -> goals(math) -> output.
Why are functional languages excellent fits for deep learning?
When a programming language expressively continues these interactions inside business logic one can control the deployments of these functions over hardware with great benefits in performance.
(fan-radio question: why is Scala a good choice for that?)
Not necessarily deep learning related. By enabling an easy way of defining an agnostic DSL Scala allows flexibility in modeling intent: ~> is beautiful.
