LBBNN

Authors Lars Skaaret-Lund, Aliaksandr Hubin, Eirik Høyheim

Last update 30 June 2026

Copyright license MIT license, File License

Software version identifier 0.1.1, 0.1.2, 0.1.3, 0.1.4, 0.1.5, PACKAGES.rds, 0.1.6

Description

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>.

Cites work

Sparse Bayesian Neural Networks: Bridging Model and Parameter Uncertainty through Scalable Variational Inference

This page was built for software: LBBNN