Create the embedding for a dataset. — make

Creates an input for each categorical var, concatenates those, Adds batch-norm to continuous vars etc.

make_embedding(task, embed_size = NULL, embed_dropout = 0)

Arguments

task	`Task` An mlr3 `Task`.
embed_size	`numeric` A numeric, either a single number determining the embedding size for all categorical features, a named vector, determining the embedding the size for each feature individually or `NULL`: Use a heuristic. The heuristic is `round(1.6 * n_cat^0.56)` where n_cat is the number of levels.
embed_dropout	`numeric` Dropout fraction for the embedding layer.

Value

A list of input tensors and layer: the concatenated embeddings.

References

Guo, Berkhan, 2016 Entity Embeddings of Categorical Variables

Examples

task = mlr3::mlr_tasks$get("boston_housing")
make_embedding(task)
#> $inputs
#> $inputs[[1]]
#> Tensor("chas:0", shape=(None, 1), dtype=int32)
#> 
#> $inputs[[2]]
#> Tensor("town:0", shape=(None, 1), dtype=int32)
#> 
#> $inputs$cont
#> Tensor("continuous:0", shape=(None, 16), dtype=float32)
#> 
#> 
#> $layers
#> Tensor("concatenate/concat:0", shape=(None, 38), dtype=float32)
#>