SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-transfer-learning-reuses-features

What does transfer learning reuse from a previously trained model?

It reuses learned features or weights for a new task.

Start playing free

  1. Only random initialization
  2. Only the old labels
  3. Previously learned representations ✔

Why: A pretrained base can supply general-purpose features instead of learning everything from scratch. A new task-specific head can then be trained on top.

Huh, didn't know that: Freezing a layer means its weights are not updated during the new training run.

The dad joke
Transfer learning packed its old features for a new trip.
Like this card?
Play the game

References

Answer: Previously learned representations. A pretrained base can supply general-purpose features instead of learning everything from scratch. A new task-specific head can then be trained on top.

  1. TensorFlow: Transfer learning and fine-tuning www.tensorflow.org/tutorials/images/transfer_learning
    take advantage of these learned feature maps without having to start from scratch
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceIn a model predicting house prices from size and bedrooms, what is the label?Artificial IntelligenceWhat kind of learning is usually used in the first, pretraining phase of an LLM?Artificial IntelligenceHow did BERT's pretraining differ from a left-to-right GPT-style model?

That's stuff everyone should know.