Embeddings

API reference: pe.embedding package.

pe.embedding.embedding.Embedding is responsible for computing the embeddings of the (synthetic or private) samples. It has the following key methods/attributes:

• pe.embedding.Embedding.compute_embedding(): Computes the embeddings of the (synthetic or private) samples.

• pe.embedding.Embedding.column_name: The column name to be used when saving the embeddings in the data_frame of pe.data.Data.

Available Embeddings

Currently, the following embeddings are implemented:

• Images

  • pe.embedding.Inception: The embeddings computed using the Inception model.

  • pe.embedding.FLDInception: The embeddings computed using the Inception model following the procedure in the FLD library.

  • pe.embedding.RawPixel: The embeddings computed by flattening the raw image pixels.

  • pe.embedding.CLIP: The embeddings computed using the CLIP model.

• Text

  • pe.embedding.SentenceTransformer: The embeddings computed using the Sentence Transformers library.

• Tabular

  • pe.embedding.TabularEmbedding: The embeddings using normalization for the numerical columns and one-hot encoding for the categorical columns.