Domain Shift
In machine learning, the phenomenon in which a model's performance degrades significantly on test data whose distribution differs from that of the training data. It is one of the core factors contributing to insufficient generalization ability in models.
Timeline (last 90 days)
The main battlefield of most general large language models is text; even with multimodal capabilities, their vision encoders are often pretrained on general image-text data and are more sensitive to domain shift in medical images.
Domain shift refers to the performance degradation that occurs when the distribution of the model's training data is inconsistent with that of the deployment data.
Domain shift is the core mechanism behind cross-institutional generalization failures in medical AI, arising from feature distribution differences caused by different hospitals using different laboratory instruments, clinical pathways, and coding practices.
The phenomenon in linguistics whereby a model's performance degrades on data outside its training distribution is called 'Domain Shift.'
The review pointed out that most models fail when externally validated, primarily due to domain shift and the absence of MLOps practices.
All Facts (5)
The main battlefield of most general large language models is text; even with multimodal capabilities, their vision encoders are often pretrained on general image-text data and are more sensitive to domain shift in medical images.
50%UnverifiedDomain shift refers to the performance degradation that occurs when the distribution of the model's training data is inconsistent with that of the deployment data.
50%UnverifiedDomain shift is the core mechanism behind cross-institutional generalization failures in medical AI, arising from feature distribution differences caused by different hospitals using different laboratory instruments, clinical pathways, and coding practices.
50%UnverifiedThe phenomenon in linguistics whereby a model's performance degrades on data outside its training distribution is called 'Domain Shift.'
50%UnverifiedThe review pointed out that most models fail when externally validated, primarily due to domain shift and the absence of MLOps practices.
50%