Massimiliano Tamborrino
@mtamborrino.bsky.social
Reader in Statistics at the University of Warwick | Inference for Stochastic processes | Simulation-based inference| Stochastic Numerics | Parallel-in-time numerical methods| Stochastic Modelling | Neuroscience | www.warwick.ac.uk/tamborrino
Thanks @spmontecarlo.bsky.social, glad you like it! There're indeed many things to check and study on both the cSMC and splitting side, with a few more on the way :) I think things would still work if without the Gaussianity and the deterministic mappings but one would need to check properly...
July 28, 2025 at 6:43 AM
Thanks @spmontecarlo.bsky.social, glad you like it! There're indeed many things to check and study on both the cSMC and splitting side, with a few more on the way :) I think things would still work if without the Gaussianity and the deterministic mappings but one would need to check properly...
It is! Especially when combined with splitting 😅
In the paper we illustrate (on a univariate cubic SDE) how Euler-Muryama-based cSMC may still explode, even when having several intermediate bridges.
Splitting-based cSMC methods don't have this problem as such...
In the paper we illustrate (on a univariate cubic SDE) how Euler-Muryama-based cSMC may still explode, even when having several intermediate bridges.
Splitting-based cSMC methods don't have this problem as such...
July 23, 2025 at 8:39 PM
It is! Especially when combined with splitting 😅
In the paper we illustrate (on a univariate cubic SDE) how Euler-Muryama-based cSMC may still explode, even when having several intermediate bridges.
Splitting-based cSMC methods don't have this problem as such...
In the paper we illustrate (on a univariate cubic SDE) how Euler-Muryama-based cSMC may still explode, even when having several intermediate bridges.
Splitting-based cSMC methods don't have this problem as such...
It has been quite a journey started a few years ago, aiming to use splitting schemes for inference, not only within ABC.
The preprint is finally out for a wide class of SDEs with non-global Lipschitz and additive diff. coefficient, fully/partially observed, with/without measurement noise!
The preprint is finally out for a wide class of SDEs with non-global Lipschitz and additive diff. coefficient, fully/partially observed, with/without measurement noise!
July 22, 2025 at 3:14 PM
It has been quite a journey started a few years ago, aiming to use splitting schemes for inference, not only within ABC.
The preprint is finally out for a wide class of SDEs with non-global Lipschitz and additive diff. coefficient, fully/partially observed, with/without measurement noise!
The preprint is finally out for a wide class of SDEs with non-global Lipschitz and additive diff. coefficient, fully/partially observed, with/without measurement noise!
Thanks @upicchini.bsky.social for the welcoming post, and thanks @marvinschmitt.com for adding me on the starting packs, you anticipated me :)
December 12, 2024 at 8:14 PM
Thanks @upicchini.bsky.social for the welcoming post, and thanks @marvinschmitt.com for adding me on the starting packs, you anticipated me :)