-
The attached code shows two cases of attempting to solve an ODE problem using ApproxFun. Changing the interval endpoint, without any other code changes, results in a shift from successful solutions t…
-
When using [Saveat](https://docs.kidger.site/diffrax/api/saveat/) we have the option to call a "fn" defined in the doc as
> fn: A function fn(t, y, args) which specifies what to save into sol.ys w…
-
With chain rule one often sees many common sub-expressions in the calculation of the function and its derivatives. It would save time to compute the two of these together rather than having one call t…
-
Let's create a worked-out example in about 1 page of code or less of an interesting computation graph that can be re-expressed as an expression graph.
The heart of PyTorch is the ability to define …
-
I want to solve a nonlinear PDE where I impose, instead of a boundary condition, that the integral over the domain is zero. I don't know how to express this – the way I tried leads to an error (see be…
-
We recently pivoted from `autograd` to `jax` for our adjoint autodifferentiation needs. While `jax` supports gradients for more functions than `autograd` it comes with a price.
`jax` relies on a s…
-
I got a couple of requests about the speed of MultiModel. It seems that for a larger network it is painfully slow. I suspect a couple of things:
- `jitcdde` compilation for large networks - but `numb…
-
Hi,
applying jax.grad to a function which uses diffrax to integrate a piecewise defined ODE, I observe that one partial derivative is unexpectedly zero. The ODE solver returns correct function values…
-
Here is an MWE:
```julia
using DifferentialEquations
using Unitful
function ode_system!(du, u, p, t)
R0, τ, Tref = p
T = u[1]*u"K"
dTdt = -T / (1 + R0*(Tref - T)) / τ
…
-
Hi devs, looks like a really nice library. I've been looking for a Jax-native root finding method that supports `vmap` for some time. Currently I am using an external call to `scipy.optimize.root` tog…