to overfit
- verb
- /ˈoʊvərˌfɪt/
- Specialized
- When developing a machine-learning model, it is crucial not to overfit it, or it may fail to perform well on new datasets.
- overfit the data
One is you can, you can do what's called overfitting to the data.
- One is you can, you can do what's called overfitting to the data.
- So I, I think I'm over applying the metaphor of a neural network in my own mind, which I just think that doesn't lead to overfitting.
Examples
-
Because the final tree overfits the data, 10-fold cross-validation is used to prune the tree.
Academic text (2012) -
Thus, it is important that when choosing a reference class, the expert does not overfit the data by including too many variables.
Academic text (2017) -
Because we should be cautious not to overfit the models due to the small sample size, we decided not to include the variable.
Academic text (2016) -
One is you can, you can do what's called overfitting to the data.
-
So the system doesn't overfit a particular evolution of the of the game of the simulation.
-
A smaller CP value leads to a bigger tree, so a smaller CP value might overfit the model to the training set.
-
While the inclusion of too few variables will result in low accuracy, the inclusion of too many variables will "overfit" the data, both resulting in a lack of predictive worth.
Academic text (2017) -
To avoid bias in our predictions, we must ensure that our model does not overfit the training data.
-
Researchers often struggle with how to prevent their algorithms from overfitting the training data while still achieving high accuracy.
Synonyms
To make a computer program learn so much from its example data that it fails on new data
Practice something so much that you only remember exact answers and cannot adapt
Antonyms
How Well
- Specialized
Surface Forms
Morphology
Etymology
The word overfit comes from over- ('too much') and fit ('to match'), just like in overcook or overdo where over- means doing too much. So to overfit a model is to make it match the data it learned from too exactly, which makes it fail on new data.