Cross validation done in whole dataset in maxent?

[shawkat sohel]

Hi All, 

Good day. Need some clarification regarding the cross-validation technique of Maxent. We know that by default cross-validation technique randomly splits the whole dataset into a number of equal-size groups called “folds”. For example, if we use a 10-fold CV that means 9 folds will be used for calibration and the left-out fold will be used for model validation. One major advantage of CV over training/test split is that it uses all of the data for validation. This ensures better use of small data sets. That means cross validation technique uses the whole dataset without splitting it into test/train. Not sure how maxent works when it uses cross validation? Because it is recommended to validate the model on unseen data. Any explanation is highly appreciated. 

Thank you.

Kind regards,

Sohel 

[shawkat sohel]

Hi All,

Can anyone please clarify my questions that I posted earlier. Need some clarification regarding the cross-validation technique of Maxent. We know that by default cross-validation technique randomly splits the whole dataset into a number of equal-size groups called “folds”. For example, if we use a 10-fold CV that means 9 folds will be used for calibration and the left-out fold will be used for model validation. One major advantage of CV over training/test split is that it uses all of the data for validation. This ensures better use of small data sets. That means cross validation technique uses the whole dataset without splitting it into test/train. Not sure how maxent works when it uses cross validation? Because it is recommended to validate the model on unseen data. Any explanation is highly appreciated. 

Thank you.