Browse Teach Curate

One-Hot Encoding

Represents a categorical value as a vector with a single 1 at that category's index. Multiplying a one-hot vector by a weight matrix selects one row, which is why an embedding layer is mathematically just a lookup.

jeremy Video

Weighted Target Encoding in Python to Prevent Data Leakage

Target encoding is a feature engineering mechanism in machine learning that transforms nominal categorical variables into continuous numeric values by substituting each category with the mean (or wei…