Guide-linked quiz · Medium Level

The Feature Hashing Trick Quiz

Covers fixed-dimensional hash buckets, vocabulary memory tradeoffs, collisions, signed hashing, and why hashed features are harder to inspect afterward.

Related guide pathsFeature Hashing Trick
Question 1 of 8

What core problem does the feature hashing trick solve for categorical or text features?

Preview every question8 questions

Answers and explanations appear as you take the quiz.

  1. What core problem does the feature hashing trick solve for categorical or text features?
    • It speeds up gradient descent convergence
    • It removes the need for any loss function
    • It avoids needing an explicit, growing dictionary to map every unique feature value to a fixed position
    • It guarantees zero hash collisions
  2. In a fixed-size FeatureHasher, how does a feature name select its output coordinate?
    • By applying a hash function to the feature's string representation and taking the result modulo the number of buckets
    • By alphabetically sorting all feature values first
    • By counting how often the feature appears in the training set
    • By asking a human annotator
  3. Two unrelated token strings land in the same output bucket. What hashing event has occurred?
    • When the hash function crashes during training
    • When a feature value has no valid hash
    • When two buckets are empty at the same time
    • When two different feature values map to the same bucket, merging their signal
  4. In signed feature hashing, what determines whether a feature contribution is added or subtracted?
    • The current training label only
    • A sign derived from the feature hash, which can be +1 or −1
    • The order in which examples arrive
    • The magnitude of the learning rate
  5. Why is the signed hashing variant generally preferred over unsigned hashing?
    • It prevents all different features from colliding
    • It stores the original feature dictionary for later recovery
    • It turns hashed vectors into dense PCA components
    • It can reduce systematic collision bias and preserve inner products in expectation under the hashing assumptions
  6. What common misconception about feature hashing does the guide directly correct?
    • That it can only be used for text data
    • That it is a form of dimensionality reduction like PCA meant to preserve variance or structure
    • That it requires labeled data
    • That it cannot be used in streaming settings
  7. Which named libraries or systems are mentioned as implementing feature hashing for streaming or out-of-core learning?
    • scikit-learn's FeatureHasher and Vowpal Wabbit
    • TensorFlow and PyTorch
    • pandas and NumPy
    • Only custom in-house tools with no public libraries
  8. In modern deep learning, what alternative to feature hashing has become common for representing categorical features, though hashing is still sometimes combined with it?
    • One-vs-rest classifiers
    • Decision tree binning
    • Learned embedding tables
    • VC dimension bounds

Keep testing yourself

More quizzes picked for your level.

View all quizzes