Expose hypergeometric_2F1 function #2792

andrjohns · 2022-07-12T08:51:02Z

Summary

This PR adds the hypergeometric_2F1 function, using the existing grad_2f1 function to calculate the gradients.

Tests

Minimal mix tests are added as the values are calculated using boost::math::hypergeometric_pFq, which is tested through the hypergeometric_pFq tests, and the gradients through grad_2F1 which already has extensive tests.

Side Effects

N/A

Release notes

Added hypergeometric_2F1 function

Checklist

Math issue Hypergeometric function naming #2664
Copyright holder: Andrew Johnson

The copyright holder is typically you or your assignee, such as a university or company. By submitting this pull request, the copyright holder is agreeing to the license the submitted work under the following licenses:
- Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
- Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
the basic tests are passing
- unit tests pass (to run, use: ./runTests.py test/unit)
- header checks pass, (make test-headers)
- dependencies checks pass, (make test-math-dependencies)
- docs build, (make doxygen)
- code passes the built in C++ standards checks (make cpplint)
the code is written in idiomatic C++ and changes are documented in the doxygen
the new changes are tested

SteveBronder

Couple small comments (and one optional big one)

stan/math/fwd/fun/hypergeometric_2F1.hpp

stan/math/prim/fun/hypergeometric_2F1.hpp

stan/math/rev/fun/hypergeometric_2F1.hpp

test/unit/math/mix/fun/hypergeometric_2F1_test.cpp

andrjohns · 2022-07-14T04:11:43Z

@SteveBronder just flagging that I'm still adding some special-case handling, so don't worry about re-reviewing just yet. Thanks!

andrjohns · 2022-07-14T20:57:24Z

@spinkey I've added a bunch of closed-form specialisations/reductions of the hypergeometric_2F1 from the WolframAlpha reference docs, and also added the Euler Transformation if the arguments don't meet convergence criteria:

$$ _2F_1(a,b;c;z) = \frac{_2F_1(c-a,b;c;\frac{z}{z-1})}{(1-z)^b} $$

Which gives a bit of a wider domain of inputs. Do you know if there are any other transformations/reductions/etc that would be good to have in?

SteveBronder

Looks good! I have a few comments for things you can optionally resolve but overall I think its clean

stan/math/prim/fun/grad_2F1.hpp

stan/math/prim/fun/grad_inc_beta.hpp

stan/math/prim/fun/hypergeometric_2F1.hpp

SteveBronder

Couple of Qs around grad_2F1_impl. I think we should either return a tuple or do in/out parameters, but I think doing both just makes things look a bit hard to read. Also I'm a little confused on why you need the user defined template parameter bool to specify whether the return should based on the partial_type_t or the return_type_t

stan/math/fwd/fun/grad_inc_beta.hpp

stan/math/prim/fun/grad_2F1.hpp

stan/math/rev/fun/grad_inc_beta.hpp

andrjohns · 2022-07-20T05:45:03Z

@SteveBronder This is ready for another look, I've tidied up the tuple/multiple-assignment handling and split out the Euler transform flow, so hopefully the codes a little more readable now

andrjohns · 2022-07-20T06:33:05Z

test/unit/math/prim/prob/wishart_cholesky_test.cpp

@@ -81,7 +81,7 @@ TEST(ProbDistributionsWishartCholesky, dof_0) {
  MatrixXd L_Y = Y.llt().matrixL();
  MatrixXd L_S = Sigma.llt().matrixL();

-  unsigned int dof = std::numeric_limits<double>::quiet_NaN();


This change was needed to fix the wishart_cholesky prim failure from this earlier Jenkins run

This was relying on undefined behavior of a NaN overflowing to 0 when assigned to an int, which looks like it didn't happen in that Jenkins run. Let me know if you'd prefer I open separate PR and issue for this

SteveBronder · 2022-07-23T14:57:42Z

Thanks!! That's fine and I'll look at this on Monday

SteveBronder

Two small optionals but looks good!

SteveBronder · 2022-07-26T15:40:15Z

stan/math/prim/fun/grad_2F1.hpp

+ * @param[in] max_steps number of steps to take
+ */
+template <bool ReturnSameT, typename T1, typename T2, typename T3, typename T_z,
+          require_not_t<std::integral_constant<bool, ReturnSameT>>* = nullptr>


[optional]

Suggested change

require_not_t<std::integral_constant<bool, ReturnSameT>>* = nullptr>

require_not_t<bool_constant<ReturnSameT>>* = nullptr>

SteveBronder · 2022-07-26T15:42:00Z

stan/math/prim/fun/grad_2F1.hpp

+template <bool ReturnSameT, typename T1, typename T2, typename T3, typename T_z,
+          require_t<std::integral_constant<bool, ReturnSameT>>* = nullptr>
+auto grad_2F1(const T1& a1, const T2& a2, const T3& b1, const T_z& z,
+              double precision = 1e-14, int max_steps = 1e6) {
+  return internal::grad_2F1_impl<true, true, true, true>(a1, a2, b1, z,
+                                                         precision, max_steps);
+}


If you set a default value for ReturnSameT like bool ReturnSameT = false do you need the other specialization below this one?

Possibly, but since the default values have to come last in the template list I'd have to make some changes to how function would get called. I might leave that as a future TODO

andrjohns and others added 3 commits July 12, 2022 10:19

Add functions initial

332dd04

Update tests and cpplint

ccb2d8a

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

1398625

SteveBronder requested changes Jul 12, 2022

View reviewed changes

andrjohns and others added 5 commits July 13, 2022 17:07

Update template docs and inlining

aca78c9

Merge branch 'develop' into add-hyper-2f1

0a5e5f5

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

b529fd1

Ensure same type scalar arguments for vector inp

64141a5

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

e054c78

andrjohns and others added 9 commits July 14, 2022 23:38

Add special case rules and prim tests

b8518d5

Begin updating grad_2f1

5139069

grad_2f1 only calculate for requested

e6f5e96

Begin updating gradients for special cases

c461bce

Update grad and hyper funs with Euler transform

30ab661

Finish doc and grad funs

0e702b2

cpplint

580508a

Merge commit 'd3b129e6783ce86b094dab62409c4eaa329d506a' into HEAD

f662efa

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

208196f

andrjohns and others added 5 commits July 15, 2022 00:22

Fix pFq 2F1 tests and doc special-cases

118a6df

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

f1b94cd

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

1b8c171

Update prim tests

8c3b188

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

3503067

SteveBronder previously approved these changes Jul 15, 2022

View reviewed changes

Review comments

884491b

andrjohns dismissed SteveBronder’s stale review via 884491b July 18, 2022 09:51

stan-buildbot and others added 2 commits July 18, 2022 05:52

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

63cd400

Update grad_pFq test args

58497c1

stan-buildbot and others added 3 commits July 18, 2022 06:30

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

74b2347

Missed 2F1 prim tests

067ec23

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

3ebb1a6

SteveBronder requested changes Jul 18, 2022

View reviewed changes

andrjohns and others added 3 commits July 19, 2022 15:53

Simplify grad_inc assignments, split euler transform logic

26e2a9a

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

99bf6f3

Fix test failure due to undefined behaviour

1475b99

andrjohns commented Jul 20, 2022

View reviewed changes

SteveBronder approved these changes Jul 26, 2022

View reviewed changes

andrjohns merged commit 2e174d2 into stan-dev:develop Jul 27, 2022

spinkney mentioned this pull request Mar 6, 2023

Expose hypergeometric_2F1 and hypergeometric_3F2 stan-dev/stanc3#1285

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Expose hypergeometric_2F1 function #2792

Expose hypergeometric_2F1 function #2792

andrjohns commented Jul 12, 2022

SteveBronder left a comment •

edited

Loading

andrjohns commented Jul 14, 2022

andrjohns commented Jul 14, 2022

SteveBronder left a comment

SteveBronder left a comment

andrjohns commented Jul 20, 2022

andrjohns Jul 20, 2022

SteveBronder commented Jul 23, 2022

SteveBronder left a comment

SteveBronder Jul 26, 2022

SteveBronder Jul 26, 2022

andrjohns Jul 27, 2022

	require_not_t<std::integral_constant<bool, ReturnSameT>>* = nullptr>
	require_not_t<bool_constant<ReturnSameT>>* = nullptr>

Expose hypergeometric_2F1 function #2792

Expose hypergeometric_2F1 function #2792

Conversation

andrjohns commented Jul 12, 2022

Summary

Tests

Side Effects

Release notes

Checklist

SteveBronder left a comment • edited Loading

Choose a reason for hiding this comment

andrjohns commented Jul 14, 2022

andrjohns commented Jul 14, 2022

SteveBronder left a comment

Choose a reason for hiding this comment

SteveBronder left a comment

Choose a reason for hiding this comment

andrjohns commented Jul 20, 2022

andrjohns Jul 20, 2022

Choose a reason for hiding this comment

SteveBronder commented Jul 23, 2022

SteveBronder left a comment

Choose a reason for hiding this comment

SteveBronder Jul 26, 2022

Choose a reason for hiding this comment

SteveBronder Jul 26, 2022

Choose a reason for hiding this comment

andrjohns Jul 27, 2022

Choose a reason for hiding this comment

SteveBronder left a comment •

edited

Loading