generalized uniform

[t4c1]

Summary

Generalizes uniform distribution related functions to accept Eigen expressions.

Tests

Tested using expression testing framework.

Side Effects

None.

Release notes

Generalized uniform distribution related functions to accept Eigen expressions.

Checklist

• Math issue Generalize matrix function signatures #1470

• Copyright holder: Tadej Ciglarič

  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

[t4c1]

Hmm, this is only failing due to tight tolerances (in test_repeat_as_vector second order gradients differ by 2.2e-14). @bbbales2 Is it possible to increase tolerances just for one function?

[bbbales2]

Yeah this seems a bit weird. I want some time to look at it.

The error went away when I changed the first lcdf test case to:

    param[0] = 1.1;   // y                                                                                                                                                   
    param[1] = -0.1;  // alpha                                                                                                                                               
    param[2] = 5.0;   // beta                                                                                                                                                
    parameters.push_back(param);
    cdf_log.push_back(std::log(1.2 / 5.1));  // expected cdf_log 

from:

    param[0] = 0.1;   // y                                                                                                                                                   
    param[1] = -0.1;  // alpha                                                                                                                                               
    param[2] = 5.0;   // beta                                                                                                                                                
    parameters.push_back(param);
    cdf_log.push_back(std::log(0.2 / 5.1));  // expected cdf_log 

It seems weird that that would have made so much of a difference for such simple functions. Anyway I will get to this after the release is sorted out.

[t4c1]

@bbbales2 Do you have time to look into this now?

[bbbales2]

@t4c1 yeah. I guess with release out no excuse to procrastinate

[bbbales2]

Alright I am confused. I was trying to simplify the computations a bit in the hopes that this went away and I am stuck.

The test I'm running is:

./runTests.py test/prob/uniform/uniform_cdf_log_00001_generated_v_test.cpp

If I replace:

ops_partials.edge1_.partials_ = inv(b_minus_a * cdf_log_n);

with:

ops_partials.edge1_.partials_ = inv(b_minus_a * (y_val - alpha_val) / b_minus_a);

Things work. If I replace it with:

ops_partials.edge1_.partials_ = inv(y_val - alpha_val);

The test fails. I have tried evaluating this expression to a temporary.

Anyway I'm not sure what is going on. I will come back to this tomorrow with a fresh head and try again.

[bbbales2]

Alright got it, but Github didn't let me push to your branch. This code passes for me:

  const auto& b_minus_a
      = to_ref_if<!is_constant_all<T_low, T_high>::value>(beta_val
                                                          - alpha_val);
  const auto& y_minus_alpha
      = to_ref_if<!is_constant_all<T_y, T_low>::value>(y_val - alpha_val);
  const auto& cdf_log_n = y_minus_alpha / b_minus_a;
  T_partials_return cdf_log = sum(log(cdf_log_n));

  if (!is_constant_all<T_y>::value) {
    if(!is_vector<T_y>::value &&
       is_vector<T_high>::value &&
       !is_vector<T_low>::value) {
      ops_partials.edge1_.partials_ = size(beta) * inv(y_minus_alpha);
    } else {
      ops_partials.edge1_.partials_ = inv(y_minus_alpha);
    }
  }
  if (!is_constant_all<T_low>::value) {
    ops_partials.edge2_.partials_
        = (y_val - beta_val) / (b_minus_a * y_minus_alpha);
  }
  if (!is_constant_all<T_high>::value) {
    if (is_vector<T_y>::value && !is_vector<T_low>::value
        && !is_vector<T_high>::value) {
      ops_partials.edge3_.partials_
          = inv(-b_minus_a) * size(y);
    } else {
      ops_partials.edge3_.partials_ = inv(-b_minus_a);
    }
  }

It replaces cdf_log_n * b_minus_a with y_minus_alpha (which is y_val - alpha_val).

[stan-buildbot]

---

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.15	3.14	1.0	0.17% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	1.0	-0.13% slower
eight_schools/eight_schools.stan	0.12	0.11	1.04	3.55% faster
gp_regr/gp_regr.stan	0.17	0.17	0.98	-1.59% slower
irt_2pl/irt_2pl.stan	5.67	5.68	1.0	-0.05% slower
performance.compilation	89.59	87.65	1.02	2.17% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.47	8.48	1.0	-0.16% slower
pkpd/one_comp_mm_elim_abs.stan	29.34	29.81	0.98	-1.61% slower
sir/sir.stan	137.92	140.48	0.98	-1.86% slower
gp_regr/gen_gp_data.stan	0.05	0.04	1.06	5.7% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.97	2.95	1.0	0.48% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.38	0.37	1.01	0.8% faster
arK/arK.stan	1.78	1.78	1.0	-0.04% slower
arma/arma.stan	0.6	0.61	0.99	-0.76% slower
garch/garch.stan	0.75	0.74	1.0	0.46% faster
Mean result: 1.00517276737

---

Jenkins Console Log
Blue Ocean
Commit hash: 86b293a

---

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

[bbbales2]

Good!