Package: LBBNN 0.1.6

LBBNN: Latent Binary Bayesian Neural Networks Using 'torch'

Latent binary Bayesian neural networks (LBBNNs) are implemented using 'torch', an R interface to the LibTorch backend. Supports mean-field variational inference as well as flexible variational posteriors using normalizing flows. The standard LBBNN implementation follows Hubin and Storvik (2024) <doi:10.3390/math12060788>, using the local reparametrization trick as in Skaaret-Lund et al. (2024) <https://openreview.net/pdf?id=d6kqUKzG3V>. Input-skip connections are also supported, as described in Høyheim et al. (2025) <doi:10.48550/arXiv.2503.10496>.

Authors:Lars Skaaret-Lund [aut, cre], Aliaksandr Hubin [aut], Eirik Høyheim [aut]

LBBNN_0.1.6.tar.gz
LBBNN_0.1.6.zip(r-4.7-any)LBBNN_0.1.6.zip(r-4.6-any)LBBNN_0.1.6.zip(r-4.5-any)
LBBNN_0.1.6.tgz(r-4.6-any)LBBNN_0.1.6.tgz(r-4.5-any)
LBBNN_0.1.6.tar.gz(r-4.7-any)LBBNN_0.1.6.tar.gz(r-4.6-any)
LBBNN_0.1.6.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION |NEWS
card.svg |card.png
LBBNN/json (API)

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

Bug tracker:https://github.com/larselund/lbbnn/issues

Datasets:

On CRAN:

Conda:

6.48 score 4 stars 274 downloads 13 exports 39 dependencies

Last updated from:1176f49950. Checks:9 OK. Indexed: yes.

TargetResultTimeFilesSyslog
linux-develOK146
source / vignettesOK358
linux-releaseOK156
macos-releaseOK171
macos-oldrelOK187
windows-develOK101
windows-releaseOK86
windows-oldrelOK94
wasm-releaseOK109

Exports:custom_activationget_dataloadersget_local_explanations_gradientlbbnn_conv2dlbbnn_linearlbbnn_netnormalizing_flowquantsresolve_devicernvp_layertorch_availabletrain_lbbnnvalidate_lbbnn

Dependencies:base64encbitbit64callrclicorocpp11descfarverggplot2gluegtableigraphisobandjsonlitelabelinglatticelifecyclemagrittrMatrixotelpkgconfigprocessxpsR6RColorBrewerRcpprlangS7safetensorsscalesstringisvglitesystemfontstextshapingtorchvctrsviridisLitewithr

classification: convolutional architecture
Demonstration of how to download KMNIST using torchvision | Create dummy dataset with the same shape as KMNIST | Create the layers that define the architecture of our convolutional network | Define the model object | Train and validate the model
Rendered fromconvolutional_architecture.Rmdusingknitr::rmarkdown

Last update: 2026-06-30
Started: 2026-06-20

classification: gallstone dataset
Prepare data | Define model | Train and validate
Rendered fromsmall_dataset_classification.Rmdusingknitr::rmarkdown

Last update: 2026-06-30
Started: 2026-06-20

Getting started with LBBNN
Introduction | Prepare dataloaders | Define the model | Train the model | Validate | Global explanation | Local explanation
Rendered fromgetting_started.Rmdusingknitr::rmarkdown

Last update: 2026-06-30
Started: 2026-06-20

variable selection: linear-data
Generate data | Define hyperparameters and the model object | Train and validate the model | Inspect the results using coef | Global structure
Rendered fromvariable_selection_linear_data.Rmdusingknitr::rmarkdown

Last update: 2026-06-30
Started: 2026-06-20

variable selection: non-linear data
Generate data | Define hyperparameters and the model object | Train and validate the model | Check the global explanations
Rendered fromvariable_selection_non_linear_data.Rmdusingknitr::rmarkdown

Last update: 2026-06-30
Started: 2026-06-20