Methods for e-means fits

Usage

# S3 method for class 'emeans'
predict(
  object,
  newdata,
  type = c("cluster", "distance"),
  device = object$recipe$device,
  ...
)

# S3 method for class 'emeans'
fitted(object, method = c("centers", "classes"), ...)

# S3 method for class 'emeans'
print(x, ...)

# S3 method for class 'emeans'
summary(object, ...)

# S3 method for class 'summary.emeans'
print(x, ...)

Arguments

object, x: An "emeans" object (x for print()).
newdata: Numeric matrix or data frame with the same columns as the training data. If both newdata and the centers have column names, the columns are matched by name. On a torch device it may also be a 2-D torch_tensor, which is cast to float32 and not checked for NA.
type: "cluster" for the nearest-center label, or "distance" for the score of that center: squared Euclidean distance, or inner-product similarity when the fit is spherical.
device: Device used to assign newdata; by default the fit's own (see "Devices" in emeans()).
...: Unused.
method: "centers" for each row's center, or "classes" for labels.

Value

predict() returns an integer vector of labels or a numeric vector of scores. fitted() returns a matrix or an integer vector. summary() returns an object of class "summary.emeans".

Examples

set.seed(1)
x <- rbind(matrix(rnorm(200, sd = 0.3), ncol = 2),
           matrix(rnorm(200, mean = 3, sd = 0.3), ncol = 2))
fit <- emeans(x, 2, threads = 1)
fit
#> e-means clustering: k = 2, 200 rows
#>   epochs 5, batch 2, alpha0 0.087
#>   tot.withinss 33.6568 (between_SS / total_SS = 96.4 %), n_splits 0
#>   device cpu
summary(fit)
#> Summary of e-means fit
#> Call: emeans(x = x, k = 2, threads = 1)
#> k = 2, epochs 5, batch 2, alpha0 0.087, device cpu
#> tot.withinss 33.6568 (between_SS / total_SS = 96.4 %), n_splits 0
#> Cluster size:
#> [1] 100 100
#> History (last rows):
#>  epoch rows      obj  time imbalance nsplit
#>      1  200 65.82473 0.011         1      0
#>      2  200 35.50273 0.021         1      0
#>      3  200 34.56189 0.032         1      0
#>      4  200 33.98544 0.043         1      0
#>      5  200 34.11096 0.057         1      0

# Label new points, or score them by their nearest center
newx <- rbind(c(0, 0), c(3, 3))
predict(fit, newx)
#> [1] 1 2
predict(fit, newx, type = "distance")
#> [1] 0.0011957318 0.0003188943

# Each training row's center, or its cluster
head(fitted(fit))
#>            [,1]        [,2]
#> [1,] 0.03266621 -0.01134242
#> [2,] 0.03266621 -0.01134242
#> [3,] 0.03266621 -0.01134242
#> [4,] 0.03266621 -0.01134242
#> [5,] 0.03266621 -0.01134242
#> [6,] 0.03266621 -0.01134242
head(fitted(fit, method = "classes"))
#> [1] 1 1 1 1 1 1