Change from split() to vector_split()

[MaurusGubser]

I use Birch for implementing a HMM for a model of two legs in a course of motion. Basically, I have a state transition model, which is linear and Gaussian with a non-diagonal covariance and an observation model, which is non-linear with Gaussian noise. Thus, I use a vector expression for the state variable, which I split up in a vector of expressions to compute the observation variable.

Previously, I used split in my code. After the most recent update of Birch, split is deprecated. I guess I should use vector_split, but it is not clear to me, how a VectorSplitExpression is used. With split, I could define a new variable x (vector of expressions), whose elements I could access using the bracket operator, see code fragment below.

function state_to_observation(u:Random<Real[_]>) -> Array<Expression<Real>> {
  let x <- split<Real>(u);
  y:Array<Expression<Real>>;

  y.insert(1, x[15]*cst_0 + x[13]*cos(x[3]) + (x[14] + g)*sin(x[3]));
  ...
  return y;
}

How can I use vector_split and how can I access the elements of the vector after using vector_split?

Not sure if this is the right place to ask a question, but I did not find any other place to do so.

[lawmurray]

Thanks for the question. The aim has been to make the use of split unnecessary in circumstances such as this, so that you can just access elements directly with a new function element(). Can you try something like this?

function state_to_observation(u:Random<Real[_]>) -> Array<Expression<Real>> {
  y:Array<Expression<Real>>;
  y.insert(1, element(u, 15)*cst_0 + element(u, 13)*cos(element(u, 3)) + (element(u, 14) + g)*sin(element(u, 3));
  ...
  return y;
}

On the to-do list is to overload the square brackets again so that e.g. element(u, 15) can be replaced with (the nicer) u[15] again.

(Footnote: No worries posting this as an issue, have been meaning to open the discussion section in GitHub, so have just done so now and moved this over there.)

[MaurusGubser]

Great, thanks a lot for your fast answer, that solved my issue. Sorry, I was not aware of the element function.

By the way, I'm really pleased working with Birch, it's a great tool.

Edit: I was too hasty. There was still an issue. Using your suggestion, I got the following error:

error: cannot convert ‘VectorElement<VectorSplitExpression<Real[_], Random<Real[_] > >, Integer>’ to ‘Expression<Real>’
src/container/Array.birch:144: note:   initializing argument 2 of ‘Array<Type>.insert(Integer, Type) [with Type = Expression<Real>; Integer = Integer]’

I could work around this using the following:

function state_to_observation(u:Random<Real[_]>) -> Array<Expression<Real>> {
  x:Array<Expression<Real>>;
  x.insert(1, box(element(u, 15))*3 + box(element(u, 13))*cos(box(element(u, 3))) + (box(element(u, 14)) + 9.81)*sin(box(element(u, 3))));
  ...
  return x;
}

But I'm not sure, if that is the intended way to do this.

[lawmurray]

You have the right idea. Here, I think you can just box the whole expression actually:

x.insert(1, box(element(u, 15)*3 + element(u, 13)*cos(element(u, 3)) + (element(u, 14) + 9.81)*sin(element(u, 3)));

What's going on here is that box creates a new Expression object each time you use it. That can get slow: objects are relatively expensive as they are heap-allocated and reference-counted. But they are also very useful, of course, as in storing them in an Array here. We've been updating lazy expressions (including the earlier change with split) to reduce the number of objects created as well as reveal many more optimization opportunities to the C++ compiler by static typing, which has produced some big performance gains recently.

[MaurusGubser]

Thanks, that worked. For now, the slower performance is not (yet) my main problem; maybe later, I need to re-think about how I implement the model to avoid creating too many objects. I'm sure, concerning performance, there is a lot to improve in my implementation.
If you're interested and it is of any benefit, I can share my model as an example, as soon as it works well.

[lawmurray]

Missed your last comment. Indeed, please share out when you're happy with it, new examples are always welcome.