Probing as Quantifying the Inductive Bias of Pre-trained Representations.
Alexander ImmerLucas Torroba HennigenVincent FortuinRyan CotterellPublished in: CoRR (2021)
Keyphrases
- pre trained
- inductive bias
- training examples
- training data
- supervised learning
- inductive learning
- machine learning
- prior knowledge
- pac learning
- small number
- training set
- learning algorithm
- hypothesis space
- control signals
- labeled data
- multitask learning
- data sets
- unlabeled data
- training samples
- action recognition
- support vector
- domain theory