Will Meierjurgen Farr

Errors in Sampling When Parameter Bounds Depend on Parameters

3

I am finding a lot of bugs in Turing samples when the bounds on distributions depend on parameters. It looks like the Jacobians of the mappings from the constrained and...

Adding Parameters to an Existing Chain

5

Perhaps this is more related to #246 but I am universally in a situation where I want to add additional parameters to my chains, and this is *unbelievably hard* in...

Order of Array Parameters is not Preserved by `get`

4

If I have a chain with array parameters, `get` returns them in some random order, not numeric: ```julia c = Chains(randn(100, 13, 4), ["x[$(i)]" for i in 13:-1:1]) xs =...

Chain / acf Plots for Ensembles Should Default to Mean Over Walkers

2

It is clearer to interpret and easier to see trends when plots like chain plots (ACF plots, etc) show the trend of the *mean position of the ensemble* versus time...

If you have data that are highly correlated or have unusual shapes, the current code in BivariateKDE does not give a very good representation of the true distribution because the...

Gradients Through Broadcasts Slow Due to Type Inference Failing?

12

I know this issue has been brought up before, but Zygote is *really slow* to take gradients through broadcasting, and I have attached a simple example that suggests (at least...

Alternative Corner Plot?

4

I'm afraid I really don't like the format of `cornerplot` when plotting the outputs of MCMC simulations. Normally I don't care so much about the correlation coefficient of my samples...