Repository navigation
Add a GPU-PSO benchmark on BioPreDyn-bench B5 - #1953
Open
AdityaPandeyCN wants to merge 1 commit into
Open
AdityaPandeyCN wants to merge 1 commit into
AdityaPandeyCN wants to merge 1 commit into
Conversation
5 tasks
AdityaPandeyCN
force-pushed
the
pso-biopredyn-b5
branch
from
October 25, 2026 07:12
bdbada5 to
61a7a5e
Compare
Fits the 86 parameters of BioPreDyn-bench B5 with the ParallelParticleSwarms optimizers and compares them with the global optimizers from the CPU B5 benchmark (BBO adaptive DE, NLopt GN_CRS2_LM) on the same wall-clock budget, 300 s per run, 3 runs each. PSOKernel and SyncPSOKernel run on the GPU with 50,000 particles, SerialPSO on the CPU with 64. The cost uses fixed-step RK4 (dt = 0.1 h) so it can run inside GPU kernels. It is checked against Rodas5P at tight tolerances and agrees to about 0.1%. Model, bounds, start point and data are copied from the CPU B5 benchmark. HybridPSO is left out. Its per-particle L-BFGS phase took 2000-2900 s with Enzyme (64 particles, limited by the tape) and ~7600 s with chunked forward mode, against the 300 s budget. ParallelParticleSwarms is pinned to b27e809 on my fork for the scalar-cost best-particle reduction and chunked ForwardDiff gradients.
AdityaPandeyCN
force-pushed
the
pso-biopredyn-b5
branch
from
October 25, 2026 07:13
61a7a5e to
2ac3f7e
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds
benchmarks/PSOBioPreDyn/pso_biopredyn_b5.jmd, a GPU-PSO version of the BioPreDyn-bench B5 parameter estimation problem (86 parameters, 26 states, 10 simulated experiments).It compares PSOKernel and SyncPSOKernel (GPU, 50,000 particles) and SerialPSO (CPU, 64 particles) with the global optimizers from the CPU B5 benchmark, BBO adaptive DE and NLopt GN_CRS2_LM. Each optimizer gets 300 s per run and 3 runs, after one untimed warm-up. The page reports the best cost at 300 s, how many runs reach the eSS reference (J = 3072.5), and a best cost vs time plot.
The cost uses fixed-step RK4 (dt = 0.1 h) instead of Rodas5P so it can run inside GPU kernels. The page checks it against Rodas5P at tight tolerances; they agree to about 0.1%.
🤖 Generated with Claude Code