Fritz Obermeyer

This is great idea! It's a near literal translation of "(stochastic) (gradient descent)".

It looks like 3.10 is being incorrectly parse as 3.1 somewhere 😕

3. Could we pose an ODE as a variational problem and solve with an argmax sum-product computation? In particular, could we have both local variables and global variables wrt a...

@eb8680 is this already completed? IIRC you did this as part of our paper submission?

cc @martinjankowiak

You'll need `ax: Bound` and `ax2: Fresh[lambda ax: ax]`. As we discussed yesterday, `@make_funsor` currently does not support binding-and-return of a dimension with the same name, so for now you'll...

For `Tensor` and `Number` you can pull out the data and re-wrap in a funsor. That is indeed hacky, and we do not have first class support for casting.

@bogiebro can you say more about your use case? Maybe give some example code? I'd like to support your casting usage. I think we could generalize the scalar type `Bint[n]`...

@eb8680 can you confirm that I can implement these as something like ```py @eager.register(Contraction, AddOp, MulOp, frozenset, Variadic[Tensor]) def eager_contract_tensor(red_op, bin_op, reduced_vars, *operands): equation = "TODO define an einsum string"...

@fehiepsi did you already complete this?