Deciding how many nodes and layers to use in a neutral network: many functions can be expressed in 1 or 2 hidden layers, more is often better and faster, and you usually have to experiment.
https://machinelearningmastery.com/how-to-configure-the-number-of-layers-and-nodes-in-a-neural-network/
miniblog.
Related Posts
Go has an elegant approach to defining example functions, which are shown in docs as `main()` with the output:
I still find it weird that constructors aren't first class functions in OCaml.
`id Just` is legal in Haskell, but `id Some` is an error in OCaml.
Are there any advantages of the OCaml approach?
Watching https://youtu.be/KWB-gDVuy_I and I'm struck by how weird constructors are as an API.
* They promote total functions, making it hard to do validation.
* They're hard to split up, because they have special access to unfinished data.
* They're like a framework: you get called.