ben18785

Interesting -- to me -- because the necessity to do this all stems from the inference problem (i.e. having different likelihoods for different outputs) I would be more tempted to...

That could work. But how would you point a particular `SubProblem` to a particular subset of outputs? Would you do: ``` collection = SharedModelProblem(forward_model, times_1, values_1, times_2, values_2) ## pertaining...

I'd also prefer to call it a `SharedProblemCollection` or (as I had above) `ProblemCollection` since it's not actually going to be of type `pints.Problem`.

I like this! ``` problem_1 = collection.subproblem(0) problem_2 = collection.subproblem(1) ```

Cool. Not sure about this one currently, to be honest. Let me think about it... On Wed, Oct 9, 2019 at 12:17 AM Michael Clerx wrote: > *@MichaelClerx* commented on...

Yeah, I like the idea! Could we also use methods from SMC to help with this? As in, we reweight particles according to their functional value, then propagate them? >...

@MichaelClerx Do we still want this? I'm not sure about it since uniformly sampling a space is easy enough to do and doesn't work well in anywhere above a handful...

I am having the same issue. The memory use increases almost monotonically even though the individual chunks are small.

Closing as this never ended up working!

List of potential discrete samplers: - [Continuous relaxation HMC](http://papers.nips.cc/paper/4652-continuous-relaxations-for-discrete-hamiltonian-monte-carlo.pdf) - Metropolis - Population - Nested(?)