density_initialization

Controls the initial density of the network, and thus an important tuning
parameter. The initialization parameters are sampled from 1/(1+exp(-U)),
with U ~ Uniform(lower,upper).

internal

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>.

Lars Skaaret-Lund

LBBNN

Latent Binary Bayesian Neural Networks Using 'torch'

Aliaksandr Hubin

Eirik Høyheim

density_initialization function

<dl><dt>lower</dt>
<dd>numeric scalar.</dd>
<dt>upper</dt>
<dd>numeric scalar, must be greater than</dd></dl>

Arguments

Generate initializations for the inclusion parameters. — density_initialization

<dl>

<dt>lower</dt>
<dd>numeric scalar.</dd>


<dt>upper</dt>
<dd>numeric scalar, must be greater than</dd>

</dl>

density_initialization: Generate initializations for the inclusion parameters.

Description

Usage

Value

Arguments