Several classes that name a learning paradigm are defined with the genus "A deep neural network that...", asserting the paradigm is-a network rather than a task or method. A reasoner will not flag this; it is a modeling choice worth revisiting.
Affected (genus = "deep neural network"):
Continual Learning, Contrastive Learning, Metric Learning, Representation Learning, Federated Learning, Deep Transfer Learning, Deep Active Learning, Zero-shot Learning, One-shot Learning, Generalized Few-shot Learning, Incremental Few-shot Learning.
Compare with sibling paradigms that already use a task genus:
- Self-supervised Learning: "A machine learning task that..."
- Transfer Learning: "A machine learning task focused on..."
Question for curators: should these be re-parented under Machine Learning Task (or Training Strategy), with any network association expressed as a separate relation rather than is-a? Surfaced during an OBO-principles (FP006) review.
Several classes that name a learning paradigm are defined with the genus "A deep neural network that...", asserting the paradigm is-a network rather than a task or method. A reasoner will not flag this; it is a modeling choice worth revisiting.
Affected (genus = "deep neural network"):
Continual Learning, Contrastive Learning, Metric Learning, Representation Learning, Federated Learning, Deep Transfer Learning, Deep Active Learning, Zero-shot Learning, One-shot Learning, Generalized Few-shot Learning, Incremental Few-shot Learning.
Compare with sibling paradigms that already use a task genus:
Question for curators: should these be re-parented under Machine Learning Task (or Training Strategy), with any network association expressed as a separate relation rather than is-a? Surfaced during an OBO-principles (FP006) review.