Implementation of Bayesian Neural Networks

[Up] [Top]

Documentation for package ‘BayesFluxR’ version 0.1.1

Help Pages

.install_pkg Installs Julia packages if needed
.julia_project_status Obtain the status of the current Julia project
.set_seed Set a seed both in Julia and R
.using Loads Julia packages
BayesFluxR_setup Set up of the Julia environment needed for BayesFlux
bayes_by_backprop Use Bayes By Backprop to find Variational Approximation to BNN.
BNN Create a Bayesian Neural Network
BNN.totparams Obtain the total parameters of the BNN
Chain Chain various layers together to form a network
Dense Create a Dense layer with 'in_size' inputs and 'out_size' outputs using 'act' activation function
find_mode Find the MAP of a BNN using SGD
Gamma Create a Gamma Prior
get_random_symbol Creates a random string that is used as variable in julia
initialise.allsame Initialises all parameters of the network, all hyper parameters of the prior and all additional parameters of the likelihood by drawing random values from 'dist'.
InverseGamma Create an Inverse-Gamma Prior
likelihood.feedforward_normal Use a Normal likelihood for a Feedforward network
likelihood.feedforward_tdist Use a t-Distribution likelihood for a Feedforward network
likelihood.seqtoone_normal Use a Normal likelihood for a seq-to-one recurrent network
likelihood.seqtoone_tdist Use a T-likelihood for a seq-to-one recurrent network.
LSTM Create an LSTM layer with 'in_size' input size, and 'out_size' hidden state size
madapter.DiagCov Use the diagonal of sample covariance matrix as inverse mass matrix.
madapter.FixedMassMatrix Use a fixed mass matrix
madapter.FullCov Use the full covariance matrix as inverse mass matrix
madapter.RMSProp Use RMSProp to adapt the inverse mass matrix.
mcmc Sample from a BNN using MCMC
Normal Create a Normal Prior
opt.ADAM ADAM optimiser
opt.Descent Standard gradient descent
opt.RMSProp RMSProp optimiser
posterior_predictive Draw from the posterior predictive distribution
prior.gaussian Use an isotropic Gaussian prior
prior.mixturescale Scale Mixture of Gaussian Prior
prior_predictive Sample from the prior predictive of a Bayesian Neural Network
RNN Create a RNN layer with 'in_size' input, 'out_size' hidden state and 'act' activation function
sadapter.Const Use a constant stepsize in mcmc
sadapter.DualAverage Use Dual Averaging like in STAN to tune stepsize
sampler.AdaptiveMH Adaptive Metropolis Hastings as introduced in
sampler.GGMC Gradient Guided Monte Carlo
sampler.HMC Standard Hamiltonian Monte Carlo (Hybrid Monte Carlo).
sampler.SGLD Stochastic Gradient Langevin Dynamics as proposed in Welling, M., & Teh, Y. W. (n.d.). Bayesian Learning via Stochastic Gradient Langevin Dynamics. 8.
sampler.SGNHTS Stochastic Gradient Nose-Hoover Thermostat as proposed in
tensor_embed_mat Embed a matrix of timeseries into a tensor
Truncated Truncates a Distribution
vi.get_samples Draw samples form a variational family.