library(tfdatasets)
data(hearts)
file <- tempfile()
writeLines(unique(hearts$thal), file)
hearts <- tensor_slices_dataset(hearts) %>% dataset_batch(32)
# use the formula interface
spec <- feature_spec(hearts, target ~ thal) %>%
step_categorical_column_with_vocabulary_list(thal) %>%
step_embedding_column(thal, dimension = 3)
spec_fit <- fit(spec)
final_dataset <- hearts %>% dataset_use_spec(spec_fit) step_embedding_column
Creates embeddings columns
Description
Use this step to create ambeddings columns from categorical columns.
Usage
step_embedding_column(
spec,
...,
dimension = function(x) {
as.integer(x^0.25)
},
combiner = "mean",
initializer = NULL,
ckpt_to_load_from = NULL,
tensor_name_in_ckpt = NULL,
max_norm = NULL,
trainable = TRUE
) Arguments
| Arguments | Description |
|---|---|
| spec | A feature specification created with feature_spec(). |
| … | Comma separated list of variable names to apply the step. selectors can also be used. |
| dimension | An integer specifying dimension of the embedding, must be > 0. Can also be a function of the size of the vocabulary. |
| combiner | A string specifying how to reduce if there are multiple entries in a single row. Currently ‘mean’, ‘sqrtn’ and ‘sum’ are supported, with ‘mean’ the default. ‘sqrtn’ often achieves good accuracy, in particular with bag-of-words columns. Each of this can be thought as example level normalizations on the column. For more information, see tf.embedding_lookup_sparse. |
| initializer | A variable initializer function to be used in embedding variable initialization. If not specified, defaults to tf.truncated_normal_initializer with mean 0.0 and standard deviation 1/sqrt(dimension). |
| ckpt_to_load_from | String representing checkpoint name/pattern from which to restore column weights. Required if tensor_name_in_ckpt is not NULL. |
| tensor_name_in_ckpt | Name of the Tensor in ckpt_to_load_from from which to restore the column weights. Required if ckpt_to_load_from is not NULL. |
| max_norm | If not NULL, embedding values are l2-normalized to this value. |
| trainable | Whether or not the embedding is trainable. Default is TRUE. |
Value
a FeatureSpec object.
Examples
See Also
steps for a complete list of allowed steps. Other Feature Spec Functions: dataset_use_spec(), feature_spec(), fit.FeatureSpec(), step_bucketized_column(), step_categorical_column_with_hash_bucket(), step_categorical_column_with_identity(), step_categorical_column_with_vocabulary_file(), step_categorical_column_with_vocabulary_list(), step_crossed_column(), step_indicator_column(), step_numeric_column(), step_remove_column(), step_shared_embeddings_column(), steps