(Varitional) ICM-EM algorithm for implementing FAST model
FAST_run(
XList,
AdjList,
q = 15,
fit.model = c("gaussian", "poisson"),
AList = NULL,
maxIter = 25,
epsLogLik = 1e-05,
verbose = TRUE,
seed = 1,
error_heter = TRUE,
Psi_diag = FALSE,
Vint_zero = FALSE
)an M-length list consisting of multiple matrices with class dgCMatrix or matrix that specifies the count/log-count gene expression matrix for each data batch used for FAST model.
an M-length list of sparse matrices with class dgCMatrix, specify the adjacency matrix used for intrisic CAR model in FAST. We provide this interface for those users who would like to define the adjacency matrix by themselves.
an optional integer, specify the number of low-dimensional embeddings to extract in FAST. Larger q means more information extracted.
an optional string, specify the version of FAST to be fitted. The Gaussian version models the log-count matrices while the Poisson verions models the count matrices; default as gaussian due to fastter computation.
an optional list with each component being a vector whose length is equal to the rows of component in XList, specify the normalization factor in FAST. The default is NULL that means the normalization factor equal to 1.
the maximum iteration of ICM-EM algorithm. The default is 30.
an optional positive vlaue, tolerance of relative variation rate of the observed pseudo loglikelihood value, defualt as '1e-5'.
a logical value, whether output the information in iteration.
a postive integer, the random seed to be set in initialization.
a logical value, whether use the heterogenous error for FAST model, default as TRUE. If error.heter=FALSE, then the homogenuous error is used.
a logical value, whether set the conditional covariance matrix of the intrisic CAR to diagonal, default as FALSE.
an optional logical value, specify whether the intial value of intrisic CAR component is set to zero; default as FALSE.
return a list including the following components: (1) hV: an M-length list consisting of spatial embeddings in FAST; (2) nu: the estimated intercept vector; (3) Psi: the estimated covariance matrix; (4) W: the estimated shared loading matrix; (5) Lam: the estimated covariance matrix of error term; (6): ELBO: the ELBO value when algorithm convergence; (7) ELBO_seq: the ELBO values for all itrations.
None
None