infer_profiles_vb: Infer methylation profiles using VB

In geneprophet/BSDMR: a region-based method for quantitating the conservation and divergence of methylation patterns between species by high order features

Description

General purpose functions for inferring latent profiles for different observation models using Variational Bayes (VB). Current observation models are: 'bernoulli', 'binomial'.

Usage

infer_profiles_vb(
  X,
  basis = NULL,
  H = NULL,
  w = NULL,
  alpha_0 = 0.5,
  beta_0 = 0.1,
  vb_max_iter = 100,
  epsilon_conv = 1e-05,
  is_parallel = TRUE,
  no_cores = NULL,
  is_verbose = FALSE,
  ...
)

Arguments

X	The input data, either a matrix or a list of elements of length N, where each element is an L X C matrix, where L are the total number of observations. The first column contains the input observations x (i.e. CpG locations). If "binomial" model then C=3, and 2nd and 3rd columns contain total number of trials and number of successes respectively.
basis	A 'basis' object. E.g. see create_rbf_object. If NULL, will an RBF object will be created.
H	Optional, design matrix of the input data X. If NULL, H will be computed inside the function.
w	A vector of initial parameters (i.e. coefficients of the basis functions). If NULL, it will be initialized inside the function.
alpha_0	Hyperparameter: shape parameter for Gamma distribution. A Gamma distribution is used as prior for the precision parameter tau.
beta_0	Hyperparameter: rate parameter for Gamma distribution. A Gamma distribution is used as prior for the precision parameter tau.
vb_max_iter	Integer denoting the maximum number of VB iterations.
epsilon_conv	Numeric denoting the convergence threshold for VB.
is_parallel	Logical, indicating if code should be run in parallel.
no_cores	Number of cores to be used, default is max_no_cores - 1.
is_verbose	Logical, print results during VB iterations.
...	Additional parameters.

Value

An object of class infer_profiles_vb_"obs_model" with the following elements:

• W: An Nx(M+1) matrix with the optimized parameter values. Each row of the matrix corresponds to each element of the list X; if X is a matrix, then N = 1. The columns are of the same length as the parameter vector w (i.e. number of basis functions).

• W_Sigma: A list with covariance matrices for each element row in W.

• basis: The basis object.

• nll_feat: NLL fit feature.

• rmse_feat: RMSE fit feature.

• coverage_feat: CpG coverage feature.

• lb_feat: Lower Bound feature.

Details

The modelling and mathematical details for inferring profiles using mean-field variational inference are explained here: http://rpubs.com/cakapourani/variational-bayes-bpr

Author(s)

Hongen Kang [e-mail]

See Also

create_rbf_object, create_region_object

Examples

# Example of inferring parameters for synthetic data using 8 RBFs
## Not run: 
basis_profile <- create_rbf_object(M=8)
human_fit_profiles <- infer_profiles_vb(X = human_obj$met, basis = basis_profile,
                                        is_parallel = TRUE, vb_max_iter = 100)

## End(Not run)

geneprophet/BSDMR documentation built on March 3, 2021, 5:50 a.m.