An ML engineer has survey answers that are only “yes” or “no.” They need to transform these responses into a single feature that improves model training without increasing the dataset’s dimensionality. Which methods satisfy these constraints? (Choose two.)

Binary encoding and Label encoding both satisfy the constraints. Binary encoding maps "yes" to 1 and "no" to 0 (or vice versa), creating a single numerical feature without increasing dimensionality. Label encoding assigns integer labels, such as 0 for "no" and 1 for "yes," achieving the same outcome. One-hot encoding would create two new columns (one for "yes" and one for "no"), increasing dimensionality, which violates a constraint. Statistical imputation is for handling missing values, not for transforming categorical data into a single feature. Tokenization is used for text processing, not for binary categorical data.