Zero-Shot Learning (ZSL)

Zero-shot learning (ZSL) is a problem setup in machine learning, where at test time, a learner observes samples from classes, which were not observed during training, and needs to predict the class that they belong to.

It’s basically Few-Shot Learning without taking any shots.

So how is that possible? Even us, humans, can’t do that.

Two ways:

  1. Pattern recognition with no training examples
  2. Solved by Semantic Transfer

I thought this was irrelevant, but actually, it’s extremely relevant. Think about all the classification tasks with a Supervised Learning paradigm. What if there are thousands of classes to predict? Some examples

  1. Object Recognition
    1. The type of butterfly
    2. The brand of a car (Toyota, Honda, Ferrari, etc.)
  2. Cross-lingual dictionary induction (?)
    1. The idea is you can translate english to german, and english to chinese. You want to translate german to chinese, wait why is this hard?
  3. Mind reading (read a brain and map that to a word)
    1. Every word would be a category

The idea is that you learn a mapping from a feature space to a category embedding.

Also, the problem of multi-label also exists.

Like say you annotate an image, maybe it has both a mountain and a beach.