Efficient Amortised Bayesian Inference for Hierarchical and Nonlinear Dynamical Systems

[MichaelClerx]

@mirams posted https://arxiv.org/abs/1905.12090

Someone should have a quick look at this

[martinjrobins]

I like this one. @ben18785, this is along the lines of what I was thinking about variational autoencoders, they look like a nice way to incorporate neural networks into inference

(Disclaimer: "along the lines of what I was thinking" means I had vague thoughts that variational autoencoders could be used to learn a posterior, definitely not saying I understand them in any meaningful way!)

[ben18785]

It looks cool but I must admit I don't really understand it. Martin, next time I'm about I'm going to quiz you about it!

…

On 12 Jun 2019, at 18:57, Martin Robinson ***@***.***> wrote: I like this one. @ben18785, this is along the lines of what I was thinking about variational autoencoders, they look like a nice way to incorporate neural networks into inference — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[martinjrobins]

Sounds good, think I'm slowly getting an understanding. The nice thing about this approach is that it gives you flexibility to define your model as an ODE, Neural Network, GP, or whatever (or a combination of a few of those). Only requirement is that it is differentiable.