Autoencoder reconstruction loss and compression (Geron Ch 18, grael)

loss is the mean squared L2 norm PER SAMPLE, mse_per_element the plain elementwise mean; compression_ratio = input_dim / code_dim.

Usage

morie_geron_autoencoder_reconstruction_loss(X, encoded, decoded)

Arguments

X

Inputs (m, d).

encoded

Codes (m, k).

decoded

Reconstructions.

Value

List with loss, mse_per_element, per_sample_loss, code_dim, input_dim, compression_ratio, explained_variance.

Examples

morie_geron_autoencoder_reconstruction_loss(X = matrix(c(1, 2, 3, 4, 5, 6), nrow = 2),
  encoded = matrix(c(1, 2, 3, 4, 5, 6), nrow = 2),
  decoded = matrix(c(1, 2, 3, 4, 5, 6), nrow = 2))
#> $loss
#> [1] 0
#> 
#> $mse_per_element
#> [1] 0
#> 
#> $per_sample_loss
#> [1] 0 0
#> 
#> $code_dim
#> [1] 3
#> 
#> $input_dim
#> [1] 3
#> 
#> $compression_ratio
#> [1] 1
#> 
#> $explained_variance
#> [1] 1
#> 
#> $estimate
#> [1] 0
#> 
#> $n
#> [1] 2
#> 
#> $method
#> [1] "Autoencoder reconstruction loss"
#>