N-Way K-Shot Terminology
What you'll learn: How to precisely describe few-shot learning scenarios using the standard N-way K-shot notation.
Understanding the Terminology
In few-shot learning, we describe the task setup using two numbers: N and K.
- N-way refers to the number of classes you need to distinguish between
- K-shot refers to the number of labeled examples you have per class
Together, "N-way K-shot" defines your support set — the tiny labeled dataset available to help the model classify new examples (the query set).
Common Examples
5-way 1-shot: You have 5 classes, with just 1 example of each class. Your support set contains only 5 labeled samples total.
3-way 5-shot: You have 3 classes, with 5 examples per class. Your support set has 15 labeled samples (3 × 5).
Real-World Analogy
Think of it like a flash-card quiz: N is how many categories you're being tested on, and K is how many example cards you get to study for each category before the test begins. If it's "4-way 2-shot," you're being quizzed on 4 types of birds, but you only got to see 2 photos of each bird species beforehand.
Why It Matters
This standardized terminology lets researchers compare approaches fairly. A model that achieves 85% accuracy on 5-way 1-shot is solving a much harder problem than one claiming 85% on 2-way 10-shot.
Key Takeaway: N-way K-shot notation precisely defines few-shot tasks: N classes with K labeled examples each in the support set, creating an NK-sample learning problem.