Zero-shot learning (ZSL) is a method of training an AI model to recognize and categorize objects or ideas without prior examples, similar to unsupervised learning. It relies on auxiliary knowledge from text data to make accurate predictions.
ZSL analyzes a substantial corpus of text with incidental references to unseen data classes. It uses this auxiliary knowledge for prediction without labeled examples, akin to human learning by reading descriptions.
[
Collection
]
[
|
...
]