Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add beta_neg_binomial_lccdf #3114

Merged

Conversation

lingium
Copy link
Collaborator

@lingium lingium commented Oct 7, 2024

Summary

With this PR the lccdf of beta negative binomial distribution are added.
See issue #3113

Expressions involved:

CCDF $C(r,\alpha,\beta)$:

$$ P(Y > y) = 1 - F(r,\alpha,\beta) = C(r,\alpha,\beta) = \frac{\Gamma (r+y +1) B(r+\alpha ,\beta +y +1) {~}_3F_2(...)}{\Gamma (r) B(\alpha ,\beta ) \Gamma (y +2)} $$

where $F(r,\alpha,\beta)$ is CDF, we denote ${~}_3F_2({1,r+y +1,\beta +y +1}; {y +2,r+\alpha +\beta +y +1};1)$ as $_3F_2(...)$

Partial derivatives:

$$ \begin{aligned} \frac{\partial \log C(r,\alpha,\beta)}{\partial r} &= \psi(r+y+1) + \psi(\alpha+r) - \psi(\alpha+\beta+r+y+1) - \psi(r) + \frac{\partial \log {}_3F_2(...)}{\partial r} \\ \frac{\partial \log C(r,\alpha,\beta)}{\partial \alpha} &= \psi(\alpha+r) - \psi(\alpha+\beta+r+y+1) + \frac{\partial \log {}_3F_2(...)}{\partial \alpha} - \psi(\alpha) + \psi(\alpha+\beta) \\ \frac{\partial \log C(r,\alpha,\beta)}{\partial \beta} &= \psi(\beta+y+1) - \psi(\alpha+\beta+r+y+1) + \frac{\partial \log {}_3F_2(...)}{\partial \beta} - \psi(\beta) + \psi(\alpha+\beta), \end{aligned} $$

where

$$ \begin{aligned} \frac{\partial \log {}_3F_2(...)}{\partial r} &= \frac{\partial {}_3F_2(...)}{\partial r} / {}_3F_2(...) \\ &= \left[ _3F_2^{({0,1,0},{0,0},0)}(...) + _3F_2^{({0,0,0},{0,1},0)}(...) \right] / _3F_2(...). \end{aligned} $$

$$ \begin{aligned} \frac{\partial \log {}_3F_2(...)}{\partial \alpha} &= {}_3F_2^{({0,0,0},{0,1},0)}(...) / {}_3F_2(...) \\ \frac{\partial \log {}_3F_2(...)}{\partial \beta} &= \left[ _3F_2^{({0,0,1},{0,0},0)}(...) + _3F_2^{({0,0,0},{0,1},0)}(...) \right] / _3F_2(...) \end{aligned} $$

These can be obtained by grad_F32.

Tests

Test is written follow the guide.

Side Effects

grad_F32 may have some precision issues, causing a few tests to fail. This has been resolved by increasing precision from 1e-6 to 1e-8.

Release notes

beta_neg_binomial_lccdf is available if merged.

Checklist

  • Copyright holder: Zhi Ling

    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

Copy link
Collaborator

@SteveBronder SteveBronder left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Few comments but overall looks good!

stan/math/prim/prob/beta_neg_binomial_lccdf.hpp Outdated Show resolved Hide resolved
stan/math/prim/prob/beta_neg_binomial_lccdf.hpp Outdated Show resolved Hide resolved
stan/math/prim/prob/beta_neg_binomial_lccdf.hpp Outdated Show resolved Hide resolved
stan/math/prim/prob/beta_neg_binomial_lccdf.hpp Outdated Show resolved Hide resolved
stan/math/prim/prob/beta_neg_binomial_lccdf.hpp Outdated Show resolved Hide resolved
Comment on lines +16 to +19
param[0] = 10; // n
param[1] = 5.5; // r
param[2] = 2.5; // alpha
param[3] = 0.5; // beta
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you add some test for more extreme values? Mostly so we can make sure we have good precision at the bounds of the support

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

will add more in the next commit

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Didn't add too many tests. Cuz if the values are extreme, grad_F32 would be too complex. We have to increase max allowed iterations to pass tests, then the tests might run for years lol

Copy link
Collaborator Author

@lingium lingium Oct 10, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

though I do test on local for all parameters = 1e-3 or 1e3, increase max_steps to 1e-8 can pass tests

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thats fine. Let's leave it at 1e-8 because we don't really know where the algorithm / user is going to end up in the parameter space.

@lingium
Copy link
Collaborator Author

lingium commented Oct 11, 2024

Hey @SteveBronder, are there any other changes you'd like to suggest?

@stan-buildbot
Copy link
Contributor


Name Old Result New Result Ratio Performance change( 1 - new / old )
arma/arma.stan 0.35 0.38 0.93 -7.23% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan 0.01 0.01 0.85 -17.68% slower
gp_regr/gen_gp_data.stan 0.03 0.03 1.01 1.38% faster
gp_regr/gp_regr.stan 0.1 0.1 1.0 0.28% faster
sir/sir.stan 70.42 71.22 0.99 -1.13% slower
irt_2pl/irt_2pl.stan 4.19 4.45 0.94 -6.26% slower
eight_schools/eight_schools.stan 0.06 0.06 0.98 -2.29% slower
pkpd/sim_one_comp_mm_elim_abs.stan 0.25 0.26 0.97 -2.62% slower
pkpd/one_comp_mm_elim_abs.stan 19.42 19.55 0.99 -0.64% slower
garch/garch.stan 0.44 0.41 1.07 6.78% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan 2.72 2.62 1.04 3.46% faster
arK/arK.stan 1.8 1.83 0.99 -1.42% slower
gp_pois_regr/gp_pois_regr.stan 2.88 3.09 0.93 -7.31% slower
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan 8.85 9.02 0.98 -1.97% slower
performance.compilation 182.31 187.06 0.97 -2.61% slower
Mean result: 0.9771050716038449

Jenkins Console Log
Blue Ocean
Commit hash: eb2b7af7527e4daaa5788c458dff4c118a19ee55


Machine information No LSB modules are available. Distributor ID: Ubuntu Description: Ubuntu 20.04.3 LTS Release: 20.04 Codename: focal

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 2400.000
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; IBRS; IBPB conditional; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

G++:
g++ (Ubuntu 9.4.0-1ubuntu1~20.04) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Clang:
clang version 10.0.0-4ubuntu1
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin

@lingium
Copy link
Collaborator Author

lingium commented Oct 21, 2024

Hi @SteveBronder, all tests passed. I think we are ready to merge :)

@SteveBronder
Copy link
Collaborator

Hi sorry I was working on a big pr for the next release and am at a conference for work right now. But Friday I'll look at this

Copy link
Collaborator

@SteveBronder SteveBronder left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Apologies for the delay. I looked this over and tried to make some changes to speed up the fvar<fvar<var>> tests. This looks good to me, but can you take a look 939f03c to make sure my changes make sense.

Otherwise we are good to go!

@stan-buildbot
Copy link
Contributor


Name Old Result New Result Ratio Performance change( 1 - new / old )
arma/arma.stan 0.38 0.34 1.14 12.08% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan 0.01 0.01 1.19 16.18% faster
gp_regr/gen_gp_data.stan 0.03 0.02 1.04 4.2% faster
gp_regr/gp_regr.stan 0.1 0.09 1.09 8.13% faster
sir/sir.stan 70.46 68.49 1.03 2.78% faster
irt_2pl/irt_2pl.stan 4.17 4.07 1.02 2.27% faster
eight_schools/eight_schools.stan 0.06 0.06 1.01 1.21% faster
pkpd/sim_one_comp_mm_elim_abs.stan 0.25 0.25 1.01 1.33% faster
pkpd/one_comp_mm_elim_abs.stan 19.48 19.4 1.0 0.39% faster
garch/garch.stan 0.44 0.41 1.07 6.91% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan 2.71 2.59 1.05 4.34% faster
arK/arK.stan 1.79 1.72 1.04 4.08% faster
gp_pois_regr/gp_pois_regr.stan 2.81 3.01 0.93 -7.02% slower
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan 8.78 8.81 1.0 -0.37% slower
performance.compilation 180.73 183.7 0.98 -1.64% slower
Mean result: 1.041382303829217

Jenkins Console Log
Blue Ocean
Commit hash: c148b0c9a356ceda75f6c6ee381d5d8ffc12bca0


Machine information No LSB modules are available. Distributor ID: Ubuntu Description: Ubuntu 20.04.3 LTS Release: 20.04 Codename: focal

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 2400.000
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; IBRS; IBPB conditional; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

G++:
g++ (Ubuntu 9.4.0-1ubuntu1~20.04) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Clang:
clang version 10.0.0-4ubuntu1
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin

@lingium
Copy link
Collaborator Author

lingium commented Oct 27, 2024

These changes make sense. Happy to go!

@lingium lingium merged commit 65fc8e6 into stan-dev:develop Oct 27, 2024
8 checks passed
@WardBrian
Copy link
Member

Note that merging this seems to have caused the develop pipeline to have gone from ~15h to a day and counting: https://jenkins.flatironinstitute.org/blue/organizations/jenkins/Stan%2FMath/detail/develop/301/

Not sure if this is repeatable or just a hiccup, we will need to monitor

@lingium
Copy link
Collaborator Author

lingium commented Oct 28, 2024

This issue is reproducible, likely due to this merge. I think the best way is to specifically adjust the beta_neg_binomial_lccdf test, (as well as the following lcdf and cdf tests later). Consider omitting or reducing the length of the RepeatAsVector and AsScalarsVsAsVector tests. The core problem is that each call to hypergeometric_3F2 (line 99, from boost. This ccdf requires the calculation of the value of hypergeometric_3F2 at z=1, where the convergence rate is very slow mathematically.) and grad_F32 (line 109) is computationally intensive, and the cumulative runtime in these tests becomes significant. Will take a look in next few days.

@lingium
Copy link
Collaborator Author

lingium commented Oct 28, 2024

Before that we can disable this test for other PRs to pass pipeline

@WardBrian
Copy link
Member

Just passed two days, I think we need to either disable that test or revert this PR until resolved

@lingium lingium deleted the feature/issue-3113-beta-neg-binomial-lccdf branch October 30, 2024 14:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants