Improve numerical stability of normal quantile gradients #3139

andrjohns · 2025-01-13T15:14:36Z

Summary

This PR updates the gradient calculations for the normal quantile functions (inv_Phi and std_normal_log_qf) to use the existing (numerically stable) standard normal density functions.

As:

$$ \frac{\text{d}}{\text{d}p} F^{-1}(p) = \frac{1}{f\left(F^{-1}(p)\right)} $$

Tests

N/A - tests should still pass

Side Effects

Improved numerical stability for gradients

Release notes

Improved the numerical stability of the gradient calculations for inv_Phi and std_normal_log_qf

Checklist

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

stan-buildbot · 2025-01-13T21:10:27Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.45	0.42	1.08	7.46% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	1.06	5.9% faster
gp_regr/gen_gp_data.stan	0.03	0.03	1.15	13.32% faster
gp_regr/gp_regr.stan	0.11	0.11	1.0	-0.16% slower
sir/sir.stan	81.53	79.4	1.03	2.61% faster
irt_2pl/irt_2pl.stan	5.44	4.63	1.17	14.85% faster
eight_schools/eight_schools.stan	0.07	0.06	1.14	11.95% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.29	0.26	1.12	10.96% faster
pkpd/one_comp_mm_elim_abs.stan	21.79	21.57	1.01	1.01% faster
garch/garch.stan	0.55	0.48	1.15	13.37% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.22	2.79	1.15	13.14% faster
arK/arK.stan	2.11	1.87	1.13	11.3% faster
gp_pois_regr/gp_pois_regr.stan	3.34	2.91	1.15	12.95% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	10.25	9.03	1.14	11.92% faster
performance.compilation	214.9	219.38	0.98	-2.08% slower
Mean result: 1.097499310152714

Jenkins Console Log
Blue Ocean
Commit hash: 318925c5ff1fb92e891ac6e599f194f520a8d465

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

stan-buildbot · 2025-01-14T20:25:11Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.34	0.41	0.83	-20.69% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	0.96	-4.36% slower
gp_regr/gen_gp_data.stan	0.03	0.03	0.98	-2.0% slower
gp_regr/gp_regr.stan	0.11	0.1	1.14	12.47% faster
sir/sir.stan	72.02	74.19	0.97	-3.02% slower
irt_2pl/irt_2pl.stan	4.23	4.55	0.93	-7.46% slower
eight_schools/eight_schools.stan	0.06	0.06	1.0	-0.47% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.25	0.26	0.94	-5.98% slower
pkpd/one_comp_mm_elim_abs.stan	19.42	20.45	0.95	-5.32% slower
garch/garch.stan	0.44	0.49	0.88	-13.01% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.73	2.86	0.95	-4.89% slower
arK/arK.stan	1.79	2.29	0.78	-27.95% slower
gp_pois_regr/gp_pois_regr.stan	2.86	3.01	0.95	-5.32% slower
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.82	8.95	0.99	-1.5% slower
performance.compilation	184.62	187.06	0.99	-1.32% slower
Mean result: 0.949398085240162

Jenkins Console Log
Blue Ocean
Commit hash: 318925c5ff1fb92e891ac6e599f194f520a8d465

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

SteveBronder

I'm fine with this as is, though I have a suggestion that I think is simpler and should work

SteveBronder · 2025-01-15T21:19:38Z

stan/math/rev/prob/std_normal_log_qf.hpp

+    auto deriv = apply_scalar_binary(
+        log_p.val(), arena_rtn, [](const auto& logp_val, const auto& rtn_val) {
+          return exp(logp_val - std_normal_lpdf(rtn_val));
+        });


Instead of apply_scalar_binary here how do you feel about just doing the std_normal_lpdf on the return val? Since then the rest can use standard vectorization from Eigen.

Suggested change

auto deriv = apply_scalar_binary(

log_p.val(), arena_rtn, [](const auto& logp_val, const auto& rtn_val) {

return exp(logp_val - std_normal_lpdf(rtn_val));

});

if constexpr (is_eigen<decltype(arena_rtn)>::value) {

auto derive = exp(log_p.val() - arena_rtn.unaryExpr([](auto x) {

return std_normal_lpdf(x);}));

log_p.adj() += elt_multiply(vi.adj(), deriv);

} else {

auto derive = exp(log_p.val() - std_normal_lpdf(arena_rtn));

log_p.adj() += vi.adj() * deriv;

}

Same thing for the other change.

andrjohns and others added 4 commits January 13, 2025 23:03

Improve numerical stability of normal quantile gradients

1e105c3

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

a67a993

Update vectorisation

b713e58

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

d7edcbd

Fix opencl headers

4f53279

SteveBronder approved these changes Jan 15, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve numerical stability of normal quantile gradients #3139

Improve numerical stability of normal quantile gradients #3139

andrjohns commented Jan 13, 2025

stan-buildbot commented Jan 13, 2025

stan-buildbot commented Jan 14, 2025

SteveBronder left a comment

SteveBronder Jan 15, 2025 •

edited

Loading

-    auto deriv = apply_scalar_binary(
-        log_p.val(), arena_rtn, [](const auto& logp_val, const auto& rtn_val) {
-          return exp(logp_val - std_normal_lpdf(rtn_val));
-        });
+    if constexpr (is_eigen<decltype(arena_rtn)>::value) {
+      auto derive = exp(log_p.val() -  arena_rtn.unaryExpr([](auto x) {
+        return std_normal_lpdf(x);}));
+      log_p.adj() += elt_multiply(vi.adj(), deriv);
+    } else {
+      auto derive = exp(log_p.val() -  std_normal_lpdf(arena_rtn));
+      log_p.adj() += vi.adj() * deriv;
+    }

Improve numerical stability of normal quantile gradients #3139

Are you sure you want to change the base?

Improve numerical stability of normal quantile gradients #3139

Conversation

andrjohns commented Jan 13, 2025

Summary

Tests

Side Effects

Release notes

Checklist

stan-buildbot commented Jan 13, 2025

stan-buildbot commented Jan 14, 2025

SteveBronder left a comment

Choose a reason for hiding this comment

SteveBronder Jan 15, 2025 • edited Loading

Choose a reason for hiding this comment

SteveBronder Jan 15, 2025 •

edited

Loading