the.com/overfitting
memorizing the answer key instead of learning the subject.
means a model that fits its training data so closely it captures noise instead of signal, then fails on anything new.
from from statistics, where fitting a curve to data becomes overfitting once you add so many parameters the curve chases every random wiggle instead of the trend.
visual testperfect on training, terrible on test data
causetoo many parameters, too little real signal
fixregularization, more data, or simpler models
human versioncramming answers instead of understanding concepts