Package: nlmixr2nn 0.0.0.9000

Matthew Fidler

nlmixr2nn: Neural-Network ODEs for 'rxode2' and 'nlmixr2'

Embed neural networks inside ordinary differential equation right-hand sides for 'rxode2' and 'nlmixr2'. Networks are exposed as compiled 'rxode2' custom functions that supply the network value and its analytic input derivatives, allowing larger networks than the per-weight code-generation approach of 'pmxNODE'.

Authors:Matthew Fidler [aut, cre]

nlmixr2nn_0.0.0.9000.tar.gz
nlmixr2nn_0.0.0.9000.zip(r-4.7-x86_64)nlmixr2nn_0.0.0.9000.zip(r-4.7-arm64)nlmixr2nn_0.0.0.9000.zip(r-4.6-x86_64)nlmixr2nn_0.0.0.9000.zip(r-4.6-arm64)nlmixr2nn_0.0.0.9000.zip(r-4.5-x86_64)
nlmixr2nn_0.0.0.9000.tgz(r-4.6-x86_64)nlmixr2nn_0.0.0.9000.tgz(r-4.6-arm64)nlmixr2nn_0.0.0.9000.tgz(r-4.5-x86_64)nlmixr2nn_0.0.0.9000.tgz(r-4.5-arm64)
nlmixr2nn_0.0.0.9000.tar.gz(r-4.7-arm64)nlmixr2nn_0.0.0.9000.tar.gz(r-4.7-x86_64)nlmixr2nn_0.0.0.9000.tar.gz(r-4.6-arm64)nlmixr2nn_0.0.0.9000.tar.gz(r-4.6-x86_64)
nlmixr2nn_0.0.0.9000.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION
card.svg |card.png
nlmixr2nn/json (API)

# Install 'nlmixr2nn' in R:
install.packages('nlmixr2nn', repos = c('https://nlmixr2.r-universe.dev', 'https://cloud.r-project.org'))

Bug tracker:https://github.com/nlmixr2/nlmixr2nn/issues

Uses libs:
  • c++– GNU Standard C++ Library v3

On CRAN:

Conda:

cpp

4.02 score 40 exports 70 dependencies

Last updated from:d2b9e8bbfe. Checks:13 WARNING, 2 OK. Indexed: yes.

TargetResultTimeFilesSyslog
linux-devel-arm64WARNING233
linux-devel-x86_64WARNING229
source / vignettesOK424
linux-release-arm64WARNING230
linux-release-x86_64WARNING250
macos-release-arm64WARNING170
macos-release-x86_64WARNING535
macos-oldrel-arm64WARNING161
macos-oldrel-x86_64WARNING334
windows-devel-arm64WARNING510
windows-devel-x86_64WARNING329
windows-release-arm64WARNING513
windows-release-x86_64WARNING280
windows-oldrel-x86_64WARNING259
wasm-releaseOK175

Exports:nnnn1nn1_d1nn1_d1_d1nn2nn2_d1nn2_d1_d1nn2_d1_d2nn2_d2nn2_d2_d2nn3nn3_d1nn3_d1_d1nn3_d1_d2nn3_d1_d3nn3_d2nn3_d2_d2nn3_d2_d3nn3_d3nn3_d3_d3nn4nn4_d1nn4_d1_d1nn4_d1_d2nn4_d1_d3nn4_d1_d4nn4_d2nn4_d2_d2nn4_d2_d3nn4_d2_d4nn4_d3nn4_d3_d3nn4_d3_d4nn4_d4nn4_d4_d4nnControlnnEvalnnnparsnnprobennWeights

Dependencies:armadillo4rbackportsBHcachemcheckmateclicpp11cpp4rcrayondata.tabledigestdparserevaluatefarverfastmapggplot2gluegtablehighrinlineisobandknitrL0Learnlabelinglatticelbfgsb3clifecyclelotrimagrittrMASSMatrixmemoiseminqamirain1qn1nanonextnlmenlmixr2datanlmixr2estnumDerivplyrPreciseSumsqs2R6RColorBrewerRcppRcppArmadilloRcppEigenRcppParallelRcppTrustreshape2rexrlangrxode2rxode2lincmtrxode2llS7scalessitmoStanHeadersstringfishstringistringrsymenginesysvctrsviridisLitewithrxfunyaml

Inside nlmixr2nn
The shape of it | Parse time: where the weights come from | Getting a value out of a UDF | The weight-block offset: the sharpest edge here | Cotangent capture | Which score is used, and why | Model augmentation | The training backend | Persistence | Things that look safe and are not | The weight penalty | Lambda is a fraction, not an amount | Extending | Where to look when something is wrong
Rendered fromnlmixr2nn-internals.Rmdusingknitr::rmarkdown

Last update: 2026-09-20
Started: 2026-08-17

Neural networks in nlmixr2 models
What nn() is | Simulating a model you have not fitted | Learning a term you do not know (UDE) | Letting the network find a covariate relationship | A model with no between-subject variability | Saving, reloading and refitting | When you want to steer the training | Regularization, when a network learns too much | What is supported
Rendered fromnlmixr2nn.Rmdusingknitr::rmarkdown

Last update: 2026-09-20
Started: 2026-07-11

Neural ODEs and nlmixr2nn: what carries over
The problem is not the same one | Sensitivities: forward, not adjoint | What that costs, and what follows from it | Regularization | The kinetic term | Augmentation, and a limit worth knowing | What is deliberately absent | References
Rendered fromnlmixr2nn-node.Rmdusingknitr::rmarkdown

Last update: 2026-09-15
Started: 2026-09-03