Owen L
Owen L
strong order is a attribute of the SDE solvers, but they also possess a weak order. Is there any reason not to include weak order as well (or should it...
We are encountering gradients that are incorrect in specific regime. Specifically, we have: - A custom solver, where the error estimate depends on a second call to the drift function...
We don't add this to the VBT because we don't know how to (to quote James "Perhaps some fancy coupling is possible here, but it's probably quite difficult."). But one...
While implementing some weak solver schemes, I noticed that when I used the VBT as opposed to the UBP (unsafe brownian path) it was substantially (like ~5 mins vs ~10...
In the docs it says "You do not need to backpropagate through the differential equation." for UBP usage. However, this doesn't seem to be theoretically necessary, you can just backprop...
A continuation of https://github.com/patrick-kidger/diffrax/pull/402 with the new 0.6.0 lineax changes. Relevant origin issue: https://github.com/patrick-kidger/diffrax/issues/401
In line with some of the weak solvers we are working on to get a PR for in diffrax, there are a variety of adaptive methods that we are implementing....
In the docs under adaptation, it says "To perform adaptive stepping with SDEs requires [commutative noise](https://docs.kidger.site/diffrax/usage/how-to-choose-a-solver/#stochastic-differential-equations). Note that this commutativity condition is not checked." Is this universally true? Can't you...
Addresses part of https://github.com/patrick-kidger/equinox/issues/810.
I previously outlined an issue in the rendering on TFQ API docs that was solved. However, it appears as though the updates did not fully solve the problem, but merely...