2.8 Ensemble Limits

BigML imposes a few limits on the characteristics of an ensemble that it can handle. Some of them are the same that applies to models.

• Number of trees: A maximum of 1,000 trees are allowed for Decision Forests and a maximum of 2,000 trees for Boosted Trees.

• Number of iterations: A maximum of 1,000 iterations are allowed for Boosted Trees.

• Fields: There is no enforced limit to the number of fields that can be present in a model.

• Instances: There is no enforced limit to the number of instances that can be handled.

• Classes: A maximum number of 1,000 distinct classes per field is allowed.

• Terms-tokens: BigML can handle up to 1,000 tokens total. In case multiple text fields are defined, then the token limit per field is divided by the number of text fields.

• Terms-full terms: BigML can handle up to 256 characters total.

• Items: A maximum of 10,000 items per field is allowed.

• Node threshold: BigML supports a value between 3 and 2,000.