Title: | Transfer Learning under Regularized Generalized Linear Models |
---|---|
Description: | We provide an efficient implementation for two-step multi-source transfer learning algorithms in high-dimensional generalized linear models (GLMs). The elastic-net penalized GLM with three popular families, including linear, logistic and Poisson regression models, can be fitted. To avoid negative transfer, a transferable source detection algorithm is proposed. We also provides visualization for the transferable source detection results. The relevant paper is available on arXiv: <arXiv:2105.14328>. |
Authors: | Ye Tian [aut, cre], Yang Feng [aut] |
Maintainer: | Ye Tian <[email protected]> |
License: | GPL-2 |
Version: | 2.0.0 |
Built: | 2024-09-12 11:02:14 UTC |
Source: | CRAN |
Fit a transfer learning generalized linear model through elastic net regularization with target data set and multiple source data sets. It also implements a transferable source detection algorithm, which helps avoid negative transfer in practice. Currently can deal with Gaussian, logistic and Poisson models.
glmtrans( target, source = NULL, family = c("gaussian", "binomial", "poisson"), transfer.source.id = "auto", alpha = 1, standardize = TRUE, intercept = TRUE, nfolds = 10, cores = 1, valid.proportion = NULL, valid.nfolds = 3, lambda = c(transfer = "lambda.1se", debias = "lambda.min", detection = "lambda.1se"), detection.info = TRUE, target.weights = NULL, source.weights = NULL, C0 = 2, ... )
glmtrans( target, source = NULL, family = c("gaussian", "binomial", "poisson"), transfer.source.id = "auto", alpha = 1, standardize = TRUE, intercept = TRUE, nfolds = 10, cores = 1, valid.proportion = NULL, valid.nfolds = 3, lambda = c(transfer = "lambda.1se", debias = "lambda.min", detection = "lambda.1se"), detection.info = TRUE, target.weights = NULL, source.weights = NULL, C0 = 2, ... )
target |
target data. Should be a list with elements x and y, where x indicates a predictor matrix with each row/column as a(n) observation/variable, and y indicates the response vector. |
source |
source data. Should be a list with some sublists, where each of the sublist is a source data set, having elements x and y with the same meaning as in target data. |
family |
response type. Can be "gaussian", "binomial" or "poisson". Default = "gaussian".
|
transfer.source.id |
transferable source indices. Can be either a subset of
|
alpha |
the elasticnet mixing parameter, with
. |
standardize |
the logical flag for x variable standardization, prior to fitting the model sequence. The coefficients are always returned on the original scale. Default is |
intercept |
the logical indicator of whether the intercept should be fitted or not. Default = |
nfolds |
the number of folds. Used in the cross-validation for GLM elastic net fitting procedure. Default = 10. Smallest value allowable is |
cores |
the number of cores used for parallel computing. Default = 1. |
valid.proportion |
the proportion of target data to be used as validation data when detecting transferable sources. Useful only when |
valid.nfolds |
the number of folds used in cross-validation procedure when detecting transferable sources. Useful only when |
lambda |
a vector indicating the choice of lambdas in transferring, debiasing and detection steps. Should be a vector with names "transfer", "debias", and "detection", each component of which can be either "lambda.min" or "lambda.1se". Component
|
detection.info |
the logistic flag indicating whether to print detection information or not. Useful only when |
target.weights |
weight vector for each target instance. Should be a vector with the same length of target response. Default = |
source.weights |
a list of weight vectors for the instances from each source. Should be a list with the same length of the number of sources. Default = |
C0 |
the constant used in the transferable source detection algorithm. See Algorithm 2 in Tian, Y. and Feng, Y., 2021. Default = 2. |
... |
additional arguments. |
An object with S3 class "glmtrans"
.
beta |
the estimated coefficient vector. |
family |
the response type. |
transfer.source.id |
the transferable souce index. If in the input, |
fitting.list |
a list of other parameters of the fitted model. |
w_athe estimator obtained from the transferring step.
delta_athe estimator obtained from the debiasing step.
target.valid.lossthe validation (or cross-validation) loss on target data. Only available when transfer.source.id = "auto"
.
source.lossthe loss on each source data. Only available when transfer.source.id = "auto"
.
thresholdthe threshold to determine transferability. Only available when transfer.source.id = "auto"
.
Tian, Y. and Feng, Y., 2021. Transfer Learning under High-dimensional Generalized Linear Models. arXiv preprint arXiv:2105.14328.
Li, S., Cai, T.T. and Li, H., 2020. Transfer learning for high-dimensional linear regression: Prediction, estimation, and minimax optimality. arXiv preprint arXiv:2006.10593.
Friedman, J., Hastie, T. and Tibshirani, R., 2010. Regularization paths for generalized linear models via coordinate descent. Journal of statistical software, 33(1), p.1.
Zou, H. and Hastie, T., 2005. Regularization and variable selection via the elastic net. Journal of the royal statistical society: series B (statistical methodology), 67(2), pp.301-320.
Tibshirani, R., 1996. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological), 58(1), pp.267-288.
predict.glmtrans
, source_detection
, models
, plot.glmtrans
, cv.glmnet
, glmnet
.
set.seed(0, kind = "L'Ecuyer-CMRG") # fit a linear regression model D.training <- models("gaussian", type = "all", n.target = 100, K = 2, p = 500) D.test <- models("gaussian", type = "target", n.target = 100, p = 500) fit.gaussian <- glmtrans(D.training$target, D.training$source) y.pred.glmtrans <- predict(fit.gaussian, D.test$target$x) # compare the test MSE with classical Lasso fitted on target data library(glmnet) fit.lasso <- cv.glmnet(x = D.training$target$x, y = D.training$target$y) y.pred.lasso <- predict(fit.lasso, D.test$target$x) mean((y.pred.glmtrans - D.test$target$y)^2) mean((y.pred.lasso - D.test$target$y)^2) # fit a logistic regression model D.training <- models("binomial", type = "all", n.target = 100, K = 2, p = 500) D.test <- models("binomial", type = "target", n.target = 100, p = 500) fit.binomial <- glmtrans(D.training$target, D.training$source, family = "binomial") y.pred.glmtrans <- predict(fit.binomial, D.test$target$x, type = "class") # compare the test error with classical Lasso fitted on target data library(glmnet) fit.lasso <- cv.glmnet(x = D.training$target$x, y = D.training$target$y, family = "binomial") y.pred.lasso <- as.numeric(predict(fit.lasso, D.test$target$x, type = "class")) mean(y.pred.glmtrans != D.test$target$y) mean(y.pred.lasso != D.test$target$y) # fit a Poisson regression model D.training <- models("poisson", type = "all", n.target = 100, K = 2, p = 500) D.test <- models("poisson", type = "target", n.target = 100, p = 500) fit.poisson <- glmtrans(D.training$target, D.training$source, family = "poisson") y.pred.glmtrans <- predict(fit.poisson, D.test$target$x, type = "response") # compare the test MSE with classical Lasso fitted on target data fit.lasso <- cv.glmnet(x = D.training$target$x, y = D.training$target$y, family = "poisson") y.pred.lasso <- as.numeric(predict(fit.lasso, D.test$target$x, type = "response")) mean((y.pred.glmtrans - D.test$target$y)^2) mean((y.pred.lasso - D.test$target$y)^2)
set.seed(0, kind = "L'Ecuyer-CMRG") # fit a linear regression model D.training <- models("gaussian", type = "all", n.target = 100, K = 2, p = 500) D.test <- models("gaussian", type = "target", n.target = 100, p = 500) fit.gaussian <- glmtrans(D.training$target, D.training$source) y.pred.glmtrans <- predict(fit.gaussian, D.test$target$x) # compare the test MSE with classical Lasso fitted on target data library(glmnet) fit.lasso <- cv.glmnet(x = D.training$target$x, y = D.training$target$y) y.pred.lasso <- predict(fit.lasso, D.test$target$x) mean((y.pred.glmtrans - D.test$target$y)^2) mean((y.pred.lasso - D.test$target$y)^2) # fit a logistic regression model D.training <- models("binomial", type = "all", n.target = 100, K = 2, p = 500) D.test <- models("binomial", type = "target", n.target = 100, p = 500) fit.binomial <- glmtrans(D.training$target, D.training$source, family = "binomial") y.pred.glmtrans <- predict(fit.binomial, D.test$target$x, type = "class") # compare the test error with classical Lasso fitted on target data library(glmnet) fit.lasso <- cv.glmnet(x = D.training$target$x, y = D.training$target$y, family = "binomial") y.pred.lasso <- as.numeric(predict(fit.lasso, D.test$target$x, type = "class")) mean(y.pred.glmtrans != D.test$target$y) mean(y.pred.lasso != D.test$target$y) # fit a Poisson regression model D.training <- models("poisson", type = "all", n.target = 100, K = 2, p = 500) D.test <- models("poisson", type = "target", n.target = 100, p = 500) fit.poisson <- glmtrans(D.training$target, D.training$source, family = "poisson") y.pred.glmtrans <- predict(fit.poisson, D.test$target$x, type = "response") # compare the test MSE with classical Lasso fitted on target data fit.lasso <- cv.glmnet(x = D.training$target$x, y = D.training$target$y, family = "poisson") y.pred.lasso <- as.numeric(predict(fit.lasso, D.test$target$x, type = "response")) mean((y.pred.glmtrans - D.test$target$y)^2) mean((y.pred.lasso - D.test$target$y)^2)
Given the point esimate of the coefficient vector from glmtrans
, calculate the asymptotic confidence interval of each component. The detailed inference algorithm can be found as Algorithm 3 in the latest version of Tian, Y. and Feng, Y., 2021. The algorithm is consructed based on a modified version of desparsified Lasso (Van de Geer, S. et al, 2014; Dezeure, R. et al, 2015).
glmtrans_inf( target, source = NULL, family = c("gaussian", "binomial", "poisson"), beta.hat = NULL, nodewise.transfer.source.id = "all", cores = 1, level = 0.95, intercept = TRUE, ... )
glmtrans_inf( target, source = NULL, family = c("gaussian", "binomial", "poisson"), beta.hat = NULL, nodewise.transfer.source.id = "all", cores = 1, level = 0.95, intercept = TRUE, ... )
target |
target data. Should be a list with elements x and y, where x indicates a predictor matrix with each row/column as a(n) observation/variable, and y indicates the response vector. |
source |
source data. Should be a list with some sublists, where each of the sublist is a source data set, having elements x and y with the same meaning as in target data. |
family |
response type. Can be "gaussian", "binomial" or "poisson". Default = "gaussian".
|
beta.hat |
initial estimate of the coefficient vector (the intercept should be the first component). Can be from the output of function |
nodewise.transfer.source.id |
transferable source indices in the infernce (the set A in Algorithm 3 of Tian, Y. and Feng, Y., 2021). Can be either a subset of
|
cores |
the number of cores used for parallel computing. Default = 1. |
level |
the level of confidence interval. Default = 0.95. Note that the level here refers to the asymptotic level of confidence interval of a single component rather than the multiple intervals. |
intercept |
whether the model includes the intercept or not. Default = TRUE. Should be set as TRUE if the intercept of |
... |
additional arguments. |
a list of output. b.hat = b.hat, beta.hat = beta.hat, CI = CI, var.est = var.est
b.hat |
the center of confidence intervals. A |
beta.hat |
the initial estimate of the coefficient vector (the same as input). |
CI |
confidence intervals (CIs) with the specific level. A |
var.est |
the estimate of variances in the CLT (Theta transpose times Sigma times Theta, in section 2.5 of Tian, Y. and Feng, Y., 2021). A |
Tian, Y. and Feng, Y., 2021. Transfer Learning under High-dimensional Generalized Linear Models. arXiv preprint arXiv:2105.14328.
Van de Geer, S., Bühlmann, P., Ritov, Y.A. and Dezeure, R., 2014. On asymptotically optimal confidence regions and tests for high-dimensional models. The Annals of Statistics, 42(3), pp.1166-1202.
Dezeure, R., Bühlmann, P., Meier, L. and Meinshausen, N., 2015. High-dimensional inference: confidence intervals, p-values and R-software hdi. Statistical science, pp.533-558.
## Not run: set.seed(0, kind = "L'Ecuyer-CMRG") # generate binomial data D.training <- models("binomial", type = "all", K = 2, p = 200) # fit a logistic regression model via two-step transfer learning method fit.binomial <- glmtrans(D.training$target, D.training$source, family = "binomial") # calculate the CI based on the point estimate from two-step transfer learning method fit.inf <- glmtrans_inf(target = D.training$target, source = D.training$source, family = "binomial", beta.hat = fit.binomial$beta, cores = 2) ## End(Not run)
## Not run: set.seed(0, kind = "L'Ecuyer-CMRG") # generate binomial data D.training <- models("binomial", type = "all", K = 2, p = 200) # fit a logistic regression model via two-step transfer learning method fit.binomial <- glmtrans(D.training$target, D.training$source, family = "binomial") # calculate the CI based on the point estimate from two-step transfer learning method fit.inf <- glmtrans_inf(target = D.training$target, source = D.training$source, family = "binomial", beta.hat = fit.binomial$beta, cores = 2) ## End(Not run)
Generate data from Gaussian, logistic and Poisson models used in the simulation part of Tian, Y. and Feng, Y., 2021.
models( family = c("gaussian", "binomial", "poisson"), type = c("all", "source", "target"), cov.type = 1, h = 5, K = 5, n.target = 200, n.source = rep(100, K), s = 5, p = 500, Ka = K )
models( family = c("gaussian", "binomial", "poisson"), type = c("all", "source", "target"), cov.type = 1, h = 5, K = 5, n.target = 200, n.source = rep(100, K), s = 5, p = 500, Ka = K )
family |
response type. Can be "gaussian", "binomial" or "poisson". Default = "gaussian".
|
type |
the type of generated data. Can be "all", "source" or "target". |
cov.type |
the type of covariates. Can be 1 or 2 (numerical). If it equals to 1, the predictors will be generated from the distribution used in Section 4.1.1 (Ah-Trans-GLM) in the latest version of Tian, Y. and Feng, Y., 2021. If it equals to 2, the predictors will be generated from the distribution used in Section 4.1.2 (When transferable sources are unknown).
|
h |
measures the deviation ( |
K |
the number of source data sets. Default = 5. |
n.target |
the sample size of target data. Should be a positive integer. Default = 100. |
n.source |
the sample size of each source data. Should be a vector of length |
s |
how many components in the target coefficient are non-zero, which controls the sparsity of target problem. Default = 15. |
p |
the dimension of data. Default = 1000. |
Ka |
the number of transferable sources. Should be an integer between 0 and |
a list of data sets which depend on the value of type
.
type
= "all": a list of two components named "target" and "source" storing the target and source data, respectively. Component source is a list containing K
components with the first Ka
ones h
-transferable and the remaining ones h
-nontransferable. The target data set and each source data set have components "x" and "y", as the predictors and responses, respectively.
type
= "source": a list with a signle component "source". This component contains a list of K
components with the first Ka
ones h
-transferable and the remaining ones h
-nontransferable. Each source data set has components "x" and "y", as the predictors and responses, respectively.
type
= "target": a list with a signle component "target". This component contains another list with components "x" and "y", as the predictors and responses of target data, respectively.
Tian, Y. and Feng, Y., 2021. Transfer Learning under High-dimensional Generalized Linear Models. arXiv preprint arXiv:2105.14328.
set.seed(0, kind = "L'Ecuyer-CMRG") D.all <- models("binomial", type = "all") D.target <- models("binomial", type = "target") D.source <- models("binomial", type = "source")
set.seed(0, kind = "L'Ecuyer-CMRG") D.all <- models("binomial", type = "all") D.target <- models("binomial", type = "target") D.source <- models("binomial", type = "source")
Plot the losses of different sources and the threshold to determine transferability for object with class "glmtrans" or "glmtrans_source_detection".
## S3 method for class 'glmtrans' plot(x, ...)
## S3 method for class 'glmtrans' plot(x, ...)
x |
an object from class "glmtrans" or "glmtrans_source_detection", which are the output of functions |
... |
additional arguments that can be passed to |
a "ggplot" visualization with the transferable threshold and losses of different sources.
Tian, Y. and Feng, Y., 2021. Transfer learning with high-dimensional generalized linear models. Submitted.
glmtrans
, source_detection
, ggplot
.
set.seed(1, kind = "L'Ecuyer-CMRG") D.training <- models("gaussian", K = 2, p = 500, Ka = 1) # plot for class "glmtrans" fit.gaussian <- glmtrans(D.training$target, D.training$source) plot(fit.gaussian) # plot for class "glmtrans_source_detection" detection.gaussian <- source_detection(D.training$target, D.training$source) plot(detection.gaussian)
set.seed(1, kind = "L'Ecuyer-CMRG") D.training <- models("gaussian", K = 2, p = 500, Ka = 1) # plot for class "glmtrans" fit.gaussian <- glmtrans(D.training$target, D.training$source) plot(fit.gaussian) # plot for class "glmtrans_source_detection" detection.gaussian <- source_detection(D.training$target, D.training$source) plot(detection.gaussian)
Predict from a "glmtrans" object based on new observation data. There are various types of output available.
## S3 method for class 'glmtrans' predict( object, newx, type = c("link", "response", "class", "integral response"), ... )
## S3 method for class 'glmtrans' predict( object, newx, type = c("link", "response", "class", "integral response"), ... )
object |
an object from class "glmtrans", which comes from the output of function |
newx |
the matrix of new values for predictors at which predictions are to be made. Should be in accordance with the data for training |
type |
the type of prediction. Default = "link". |
... |
additional arguments.
|
the predicted result on new data, which depends on type
.
Tian, Y. and Feng, Y., 2021. Transfer learning with high-dimensional generalized linear models. Submitted.
set.seed(1, kind = "L'Ecuyer-CMRG") # fit a logistic model D.training <- models("binomial", type = "all", K = 1, p = 500) D.test <- models("binomial", type = "target", n.target = 10, p = 500) fit.binomial <- glmtrans(D.training$target, D.training$source, family = "binomial") predict(fit.binomial, D.test$target$x, type = "link") predict(fit.binomial, D.test$target$x, type = "response") predict(fit.binomial, D.test$target$x, type = "class") # fit a Poisson model D.training <- models("poisson", type = "all", K = 1, p = 500) D.test <- models("poisson", type = "target", n.target = 10, p = 500) fit.poisson <- glmtrans(D.training$target, D.training$source, family = "poisson") predict(fit.poisson, D.test$target$x, type = "response") predict(fit.poisson, D.test$target$x, type = "integral response")
set.seed(1, kind = "L'Ecuyer-CMRG") # fit a logistic model D.training <- models("binomial", type = "all", K = 1, p = 500) D.test <- models("binomial", type = "target", n.target = 10, p = 500) fit.binomial <- glmtrans(D.training$target, D.training$source, family = "binomial") predict(fit.binomial, D.test$target$x, type = "link") predict(fit.binomial, D.test$target$x, type = "response") predict(fit.binomial, D.test$target$x, type = "class") # fit a Poisson model D.training <- models("poisson", type = "all", K = 1, p = 500) D.test <- models("poisson", type = "target", n.target = 10, p = 500) fit.poisson <- glmtrans(D.training$target, D.training$source, family = "poisson") predict(fit.poisson, D.test$target$x, type = "response") predict(fit.poisson, D.test$target$x, type = "integral response")
Similar to the usual print methods, this function summarizes results from a fitted "glmtrans"
object.
## S3 method for class 'glmtrans' print(x, ...)
## S3 method for class 'glmtrans' print(x, ...)
x |
fitted |
... |
additional arguments. |
No value is returned.
set.seed(1, kind = "L'Ecuyer-CMRG") # fit a linear model D.training <- models("gaussian", K = 2, p = 500) fit.gaussian <- glmtrans(D.training$target, D.training$source) fit.gaussian
set.seed(1, kind = "L'Ecuyer-CMRG") # fit a linear model D.training <- models("gaussian", K = 2, p = 500) fit.gaussian <- glmtrans(D.training$target, D.training$source) fit.gaussian
Detect transferable sources from multiple source data sets. Currently can deal with Gaussian, logistic and Poisson models.
source_detection( target, source = NULL, family = c("gaussian", "binomial", "poisson"), alpha = 1, standardize = TRUE, intercept = TRUE, nfolds = 10, cores = 1, valid.nfolds = 3, lambda = "lambda.1se", detection.info = TRUE, target.weights = NULL, source.weights = NULL, C0 = 2, ... )
source_detection( target, source = NULL, family = c("gaussian", "binomial", "poisson"), alpha = 1, standardize = TRUE, intercept = TRUE, nfolds = 10, cores = 1, valid.nfolds = 3, lambda = "lambda.1se", detection.info = TRUE, target.weights = NULL, source.weights = NULL, C0 = 2, ... )
target |
target data. Should be a list with elements x and y, where x indicates a predictor matrix with each row/column as a(n) observation/variable, and y indicates the response vector. |
source |
source data. Should be a list with some sublists, where each of the sublist is a source data set, having elements x and y with the same meaning as in target data. |
family |
response type. Can be "gaussian", "binomial" or "poisson". Default = "gaussian".
|
alpha |
the elasticnet mixing parameter, with
. |
standardize |
the logical flag for x variable standardization, prior to fitting the model sequence. The coefficients are always returned on the original scale. Default is |
intercept |
the logical indicator of whether the intercept should be fitted or not. Default = |
nfolds |
the number of folds. Used in the cross-validation for GLM elastic net fitting procedure. Default = 10. Smallest value allowable is |
cores |
the number of cores used for parallel computing. Default = 1. |
valid.nfolds |
the number of folds used in cross-validation procedure when detecting transferable sources. Useful only when |
lambda |
lambda (the penalty parameter) used in the transferable source detection algorithm. Can be either "lambda.min" or "lambda.1se". Default = "lambda.1se". |
detection.info |
the logistic flag indicating whether to print detection information or not. Useful only when |
target.weights |
weight vector for each target instance. Should be a vector with the same length of target response. Default = |
source.weights |
a list of weight vectors for the instances from each source. Should be a list with the same length of the number of sources. Default = |
C0 |
the constant used in the transferable source detection algorithm. See Algorithm 2 in Tian, Y. and Feng, Y., 2021. Default = 2.
|
... |
additional arguments. |
An object with S3 class "glmtrans_source_detection"
.
target.valid.loss |
the validation (or cross-validation) loss on target data. Only available when |
source.loss |
the loss on each source data. Only available when |
threshold |
the threshold to determine transferability. Only available when |
source.loss
and threshold
outputed by source_detection
can be visualized by function plot.glmtrans
.
Tian, Y. and Feng, Y., 2021. Transfer Learning under High-dimensional Generalized Linear Models. arXiv preprint arXiv:2105.14328.
Li, S., Cai, T.T. and Li, H., 2020. Transfer learning for high-dimensional linear regression: Prediction, estimation, and minimax optimality. arXiv preprint arXiv:2006.10593.
Friedman, J., Hastie, T. and Tibshirani, R., 2010. Regularization paths for generalized linear models via coordinate descent. Journal of statistical software, 33(1), p.1.
Zou, H. and Hastie, T., 2005. Regularization and variable selection via the elastic net. Journal of the royal statistical society: series B (statistical methodology), 67(2), pp.301-320.
Tibshirani, R., 1996. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological), 58(1), pp.267-288.
glmtrans
, predict.glmtrans
, models
, plot.glmtrans
, cv.glmnet
, glmnet
.
set.seed(0, kind = "L'Ecuyer-CMRG") # study the linear model D.training <- models("gaussian", type = "all", K = 2, p = 500, Ka = 1, n.target = 100, cov.type = 2) detection.gaussian <- source_detection(D.training$target, D.training$source) detection.gaussian$transferable.source.id # study the logistic model D.training <- models("binomial", type = "all", K = 2, p = 500, Ka = 1, n.target = 100, cov.type = 2) detection.binomial <- source_detection(D.training$target, D.training$source, family = "binomial", cores = 2) detection.binomial$transferable.source.id # study Poisson model D.training <- models("poisson", type = "all", K = 2, p = 500, Ka = 1, n.target = 100, cov.type = 2) detection.poisson <- source_detection(D.training$target, D.training$source, family = "poisson", cores = 2) detection.poisson$transferable.source.id
set.seed(0, kind = "L'Ecuyer-CMRG") # study the linear model D.training <- models("gaussian", type = "all", K = 2, p = 500, Ka = 1, n.target = 100, cov.type = 2) detection.gaussian <- source_detection(D.training$target, D.training$source) detection.gaussian$transferable.source.id # study the logistic model D.training <- models("binomial", type = "all", K = 2, p = 500, Ka = 1, n.target = 100, cov.type = 2) detection.binomial <- source_detection(D.training$target, D.training$source, family = "binomial", cores = 2) detection.binomial$transferable.source.id # study Poisson model D.training <- models("poisson", type = "all", K = 2, p = 500, Ka = 1, n.target = 100, cov.type = 2) detection.poisson <- source_detection(D.training$target, D.training$source, family = "poisson", cores = 2) detection.poisson$transferable.source.id