ben18785
ben18785
At the moment, it's a really bizarre model because: - It only outputs infected and recovered populations (even though we don't observe the susceptible population, we should still output it)....
[Numpyro](https://github.com/pyro-ppl/numpyro) is a lightweight probabilistic programming language allowing automatic differentiation via Jax and JIT compilation to GPU/TPU/CPU. We would like to potentially build an interface to numpyro but, first, it'd...
Only requires a few extra lines on top of existing algorithms, they claim [here](https://projecteuclid.org/euclid.ba/1572055334).
This is a slightly self-involved one but I do actually think it'd be useful. I would like to add my R* convergence metric that I describe in [this](https://arxiv.org/abs/2003.07900) paper, as...
Allows diagnosis of regions of high curvature.
To begin with we should probably state that PINTS does not allow discrete parameters. Later on, we should probably implement some discrete samplers - for some of the samplers, these...
We should compare PINTS' outputs for its implemented samplers with those of Stan (or Pymc3) -- the current industry standard. In the future of PINTS development, I think it's important...
At the moment the series of intermediate distributions are different between SMC and populationMCMC. We should allow each to have the other's option and make the default the same between...
Not sure how easy this is to do in Pints but will likely be very useful for bigger models, with correlated parameters.
Another really interesting direction for Pints to go in is to look at methods for sampling from intractable (i.e. unlike the Gaussian problem that are solved by use of Kalman...