One-Hot Encoding (Multi-class)

Feature EngineeringData Processing

Medium

Convert $N$ integer class labels into an $N\times K$ one-hot matrix. Row $i$ is defined by:

Y_{ij} = \begin{cases}1, & j=y_i \\ 0, & j\ne y_i\end{cases}

Here, $y_i$ is the label for sample $i$ and $K$ is num_classes. When num_classes is None, set $K=\max(y)+1$ . Return a floating-point NumPy array.

Loading visualization...

Examples

Input: y = [0, 2, 1], num_classes = None

Output: [[1.0, 0.0, 0.0], [0.0, 0.0, 1.0], [0.0, 1.0, 0.0]]

Explanation: The largest label implies three columns, and each row marks its label index.

Input: y = [1, 1, 0], num_classes = 4

Output: [[0.0, 1.0, 0.0, 0.0], [0.0, 1.0, 0.0, 0.0], [1.0, 0.0, 0.0, 0.0]]

Hint 1

Create np.zeros((y.size, num_classes), dtype=float).

Hint 2

Assign ones with encoded[np.arange(y.size), y] = 1.0.

Requirements

Infer the class count when num_classes is None
Assign one active entry per row with vectorized indexing
Return a floating-point NumPy array of shape (N,K)

Constraints

y is a nonempty one-dimensional list of nonnegative integer labels
num_classes is None or greater than the largest label
Use NumPy only

Try Similar Problems

Minmax Normalization Streaming Minmax Ordinal Encoding Target Encoding Frequency Encoding

Prev Next