liveBook · Manning

CategoryEncoding

As explained in Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI

CategoryEncoding is a Keras preprocessing layer that converts integer category indices into encoded representations.

CategoryEncoding converts integer category indices into encoded representations, typically completing a categorical preprocessing pipeline after StringLookup or IntegerLookup. section 9.3.2

Listing 9.13 Creating encoding layers for categorical columns
# function from 
https://www.tensorflow.org/tutorials/structured_data/preprocessing_layers
def get_category_encoding_layer(name, dataset, dtype, max_tokens=None):
  if dtype == 'string':                                     #1
    index = layers.StringLookup(max_tokens=max_tokens)
  else:
    index = layers.IntegerLookup(max_tokens=max_tokens)
  feature_ds = dataset.map(lambda x, y: x[name])            #2
  index.adapt(feature_ds)                                   #3
  encoder = \
layers.CategoryEncoding(num_tokens= \
index.vocabulary_size())                                    #4
  return lambda feature: encoder(index(feature))            #5

Build the encoder by adapting a lookup layer to a dataset, sizing CategoryEncoding from the vocabulary, and composing both transformations. section 9.3.2

Encoding real-estate categories

A Kuala Lumpur property model gives each categorical column an input, lookup layer, and CategoryEncoding layer before concatenating encoded features with normalized continuous columns. One-hot encoding fits columns with roughly 100 or fewer unique values, while very large vocabularies may require embeddings to avoid memory problems. section 9.3.2

How Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI covers it

How deep it goes
Shows dataset adaptation, vocabulary-sized encoding, and pipeline composition, but defers CategoryEncoding internals.
Examples
Uses Location, Property Type, Furnishing, and Size_type_bin from a Kuala Lumpur real-estate dataset.
What it assumes
Assumes tabular datasets and categorical columns, while demonstrating lookup adaptation and feature-pipeline assembly.
What you can do after
Build it from scratch by adapting a lookup layer and wrapping it with a vocabulary-sized CategoryEncoding layer.
How it connects
Connects categorical inputs with StringLookup, IntegerLookup, one-hot encoding, embeddings, and Keras model layers.
Cover of Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
liveBookmanning.com
Read Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI on liveBook Get the book at manning.com

This explainer was generated by Manning's knowledge graph from the text of Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI. The section links point to the parts of the book it draws on.