Zero-Shot Learning
Zero-Shot Learning is a learning paradigm in machine learning in which a model can recognize or reason about target categories without having encountered any training samples for those categories. Its core mechanism involves using semantic descriptions, attribute vectors, or knowledge transfer to generalize knowledge from known categories to unknown ones. This approach is widely applied in image recognition and natural language processing tasks, and represents an important technique for addressing the problem of scarce labeled data.
Core Facts
Timeline (last 90 days)
Zero-shot learning is typically achieved via two pathways: meta-learning and generalization based on structural priors.
Zero-shot Learning refers to a model's ability to complete new tasks solely based on general knowledge accumulated during the pre-training phase, without having been trained on those specific tasks.
CLIP achieves zero-shot classification, recognizing categories that were never explicitly seen during training.
TimesFM has zero-shot forecasting capabilities, enabling it to directly forecast unseen time series without fine-tuning.
One-shot learning is not the same as zero-shot learning; the single demonstration relies on prior knowledge accumulated by the model from massive pretraining data.
All Facts (5)
One-shot learning is not the same as zero-shot learning; the single demonstration relies on prior knowledge accumulated by the model from massive pretraining data.
65%UnverifiedZero-shot learning is typically achieved via two pathways: meta-learning and generalization based on structural priors.
50%UnverifiedZero-shot Learning refers to a model's ability to complete new tasks solely based on general knowledge accumulated during the pre-training phase, without having been trained on those specific tasks.
50%UnverifiedCLIP achieves zero-shot classification, recognizing categories that were never explicitly seen during training.
50%UnverifiedTimesFM has zero-shot forecasting capabilities, enabling it to directly forecast unseen time series without fine-tuning.
50%