predict.catch {TULIP}

Predict categorical responses for matrix/tensor data.

Description

Predict categorical responses on new matrix/tensor data given the fitted CATCH model input.

Usage

## S3 method for class 'catch'
predict(object, newx, z = NULL, ztest = NULL, gamma = NULL,...)

Arguments

object

Input catchobj class object as fitted model.

newx

Input adjusted testing tensor or matrix list. Each element of the list is a tensor. The tensor should of the same dimension as training data.

z

Input training covariates matrix. z can be omitted if there is no covariate.

ztest

Input testing covariates matrix. ztest can be omitted if there is no covariate.

gamma

Coefficients of covariates obtained from adjten. gamma is NULL if there is no covariate.

...

Other arguments that can be passed to predict.

Details

The function fits LDA model on selected discriminant vectors. Call predict or predict.catch to perform predictions.

There are two ways to make predictions. One way is to directly predict at the same time as fitting model by catch since predict.catch has already been embedded in catch and it will predicts response when testing data is provided. The other way is to first use adjten to adjuste tensor and catch to fit model. predict.catch will take the input adjusted tensor list newx, covariate coefficient gamma from adjten and the fitted model from catch to perform prediction. The prediction is identical to providing catch testing data.

Value

Predicted response of newx for each lambda in model object.

Author(s)

Yuqing Pan, Qing Mai, Xin Zhang

References

Pan, Y., Mai, Q., and Zhang, X. (2018) Covariate-Adjusted Tensor Classification in High-Dimensions, arXiv:1805.04421.

See Also

catch, adjten

Examples

#generate training data
n <- 20
p <- 4
k <- 2
nvars <- p*p*p
x <- array(list(),n)
vec_x <- matrix(rnorm(n*nvars),nrow=n,ncol=nvars)
vec_x[1:10,] <- vec_x[1:10,]+2
z <- matrix(rnorm(n*2),nrow=n,ncol=2)
z[1:10,] <- z[1:10,]+0.5
y <- c(rep(1,10),rep(2,10))
for (i in 1:n){
  x[[i]] <- array(vec_x[i,],dim=c(p,p,p))
}

#generate testing data
newx <- array(list(),n)
vec_newx <- matrix(rnorm(n*nvars),nrow=n,ncol=nvars)
vec_newx[1:10,] <- vec_newx[1:10,]+2
newz <- matrix(rnorm(n*2),nrow=n,ncol=2)
newz[1:10,] <- newz[1:10,]+0.5
for (i in 1:n){
  newx[[i]] <- array(vec_newx[i,],dim=c(p,p,p))
}

#Make adjustment and fit model
obj <- adjten(x, z, y, newx, newz)
fit <- catch(x, z, y)
#Predict
pred <- predict(fit, obj$testxres, z, newz, obj$gamma)

#The adjusting, fitting model and predicting step can also be completed
#by one command.
pred <- catch(x, z, y, newx, newz)$pred