Gaurav Arya

Hi, thank you for the report:) It's a bit of a hectic time, so I just wanted to let you know that it may be a few weeks before I...

Ah, you could try registering your deterministic model as a single StochasticAD primitive via https://gaurav-arya.github.io/StochasticAD.jl/dev/devdocs.html#via-StochasticAD.propagate and see if that yields any speedup 🙂

A new issue for that would definitely be appropriate! I don't know much about GPUs, but my guess is that it would be important to write rules for vector operations...

Your implementation above, with concurrent discrete random steps for S I and R, should be unproblematic for stochastic triples 🙂

Sorry, I was looking at just your small snippet above, not the full particle filter you originally sent. In #70 I've added `new_weight` support for stochastic triples, so that the...

Hey! Sorry for the late reply here. If you're interested, I think one interesting direction to go in would be to push through the smoothing approach to work on the...

Hey! First, would you mind reporting the [standard error](https://en.wikipedia.org/wiki/Standard_error) of your Monte Carlo measurements along with the means? (This should equal to the standard deviation of the samples divided by...

```julia function rand_walk(STATE_CHAIN,RATE_VEC,END_VEC,T) p = 1-exp(-T*RATE_VEC[STATE_CHAIN[1]]) # probability of jump B = rand(Bernoulli(p))+1 # 1 is no jump, 2 is jump # rng for exponential conditioned to jump before T...

The reason `StochasticAD` fails is because the `t` contains a dual component which is dropped in the `t < T` comparison. (See the fifth bullet point of https://gaurav-arya.github.io/StochasticAD.jl/stable/limitations.html) One simple...

Hey! Although I haven't gotten around to explicitly handling `ifelse`, there is a more powerful (experimental) `StochasticAD.propagate` function: see the branching example at https://gaurav-arya.github.io/StochasticAD.jl/stable/devdocs.html#via-StochasticAD.propagate