FP-Syd meeting!

FP Syd is on again, for the last time in 2023[1]. We have Jonathan Merritt speaking, who will enlighten us about automatic differentiation using `backprop` (abstract below).

We will again reserve some time for short lightning talks before the main talk of this month. If you had a great idea recently, if you have worked with a cool library you can recommend, or if you want to present something funny... the stage is yours.

As usual, doors open at 6pm for the talks to start around 6:45.

We are always on the lookout for new (and returning!) speakers. If you'd like to present something Functional Programming related next year, get in touch with the organisers or open a PR on http://github.com/fp-syd/meetings.

[1] We usually skip the December meeting, and will be back on 24 January 2024 after this edition.

------------------------------------------------------------
Jonathan Merritt: A brief tour of the `backprop` library in Haskell

The `backprop` library does reverse-mode automatic differentiation in Haskell. In this talk, I’ll cover how reverse-mode AD computes derivatives, and show some examples of using `backprop` with gradient descent to solve optimisation problems.
One of the promises of this kind of “differentiable programming” is that we can compute gradients for pure Haskell functions that use generic type classes, like `Num` or `VectorSpace`, that aren’t specifically written against the types used in a neural network library. I’ll show an example of doing backprop through a Runge-Kutta ODE solver that demonstrates this.