Introduction to machine learning · Question 49 of 60
A classifier for apple varieties is to be adapted by fine-tuning. Which target tasks are likely to work well and which less well?
2 answers are correct. Hard
Show answer
Answer
- A Well: recognizing pear varieties in photos
- B Less well: recognizing bird songs
Explanation
Fine-tuning works better the closer the new task is to the old one. From apple photos to pear photos is a small step, and the model can use much of what it already knows. Bird songs are audio recordings, where what it has learned hardly helps. For the same reason, an LLM can be fine-tuned well for a closely related specialist task if good training data is available.