What is the actual meaning of these two points in Big Query. I get that in 2nd, maybe, by total cardinality it actually means no. of features. What about point 1?

1. If total cardinalities of training features are more than 10,000, batch_gradient_descent strategy is used.
2. If there is over-fitting issue, that is, the number of training examples is less than 10x, where x is total cardinality, batch_gradient_descent strategy is used.

CodePudding user response：

The cardinality is the number of possible values for a feature. The total is the sum of the possible values of all the features.

For the #2, that means you must provide at least 10 time more input than the sum of the possible values of all the feature. That is for ensuring that you have enough examples per cardinality and therefore prevent over-fitting.