ben18785

It's standard practice to randomly initialise start points for sampling. In most notebooks, we don't do this: we just set parameters deterministically. I'd like to change this so that, in...

Set temperature schedule on population MCMC to be uniform on log scale

5

As per SMC (and per Radford Neal's AIS article on p10: ftp://www.cs.toronto.edu/pub/radford/ais-rev.pdf)

Is there a need for SingleOutputProblem and MultiOutputProblem?

1

Both methods take the same arguments. Indeed, looking at the two classes, they're both virtually identical. They both use `model.n_parameters()` to check that the user has provided an object of...

Add more useful log likelihoods

10

There are a number of log-likelihoods which would be useful (this came up in discussion for the [PKPD app](https://github.com/pkpdapp-team/pkpdapp).): - [x] log-normal: used a lot in PKPD modelling and elsewhere...

[This](https://link.springer.com/article/10.1007/s11222-006-9438-0) paper has a few examples: - example 2 is a simple chemical reaction equation that is non-identifiable (so good for illustrating use of fake data simulations) - example 3...

Tried running DramACMC on `pints.GaussianLogPDF` and it looks like the distribution converges to a stationary distribution that is substantially wider than the target.

Matplotlib looks like it has now added error checking for the value of `alpha` (which controls the transparency of plots) as I now get the following types of error when...

Add evaluateS1 method for HighDimensionalGaussianLogPDF

2

Why? Because it'll be very useful for functionally testing gradient-based methods. I think this can just be copied straight over from that for `GaussianLogPDF`. Whilst we're at it, it might...