Regularization relaxation scheme

1. A computer-implemented method for training a machine learning model, the method comprising:

receiving, by a machine learning system, a plurality of training examples, wherein each training example comprises one or more features that identify properties of a respective training instance;

training, by the machine learning system, the machine learning model on the plurality of training examples to determine trained values for weights of the machine learning model, wherein the machine learning model comprises a respective weight for each feature found in any of the training examples processed by the machine learning model, wherein, during the training, the machine learning model processes each of the training examples to determine a predicted output for the training example from current values of the weights for the features in the training example, and wherein training the machine learning model further comprises:

assigning a respective initial value for a regularization penalty for a particular feature, wherein the machine learning model determines the predicted output for each training example that comprises the particular feature by performing computations that include combining a current value for the regularization penalty for the particular feature with the current value of the particular weight for the particular feature, and

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the particular feature and (2) have a label specifying a particular outcome, exceeds a threshold frequency, and, if so, decreasing the current value for the regularization penalty for the particular feature during the training of the machine learning model; and

generating, after completing the training of the machine learning model, a trained machine learning model based on the trained values for the weights whose values have been adjusted based on the respective regularization penalties.

2. The computer-implemented method of claim 1 , wherein training the machine learning model further comprises:

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the particular feature and (2) have a label specifying a particular outcome, does not exceed a threshold frequency, and, if so, increasing the current value for the regularization penalty for the particular feature during the training of the machine learning model.

3. The computer-implemented method of claim 1 , wherein the particular feature belongs to a base template, wherein a base template is a feature category that includes features of a single feature type.

4. The computer-implemented method of claim 1 , wherein the particular feature belongs to a cross-template, wherein a cross-template is a feature category that includes a plurality of features, with each of the plurality of features being of a respective feature type.

5. The computer-implemented method of claim 1 , wherein combining a current value for the regularization penalty for the particular feature with the current value of the particular weight for the particular feature, includes dividing the current value of the particular weight for the particular feature by the current value for the regularization penalty for the particular feature.

6. The computer-implemented method of claim 1 , wherein the training examples are potential video viewings, and wherein the predicted output is a prediction of whether a user will view a particular video.

7. The computer-implemented method of claim 1 , wherein training the machine learning model further comprises:

for each feature found in any of the training examples:

assigning a respective initial value for a regularization penalty for the feature, wherein the machine learning model determines the predicted output for each training example that comprises the feature by performing computations that include combining a current value for the regularization penalty for the feature with the current value of the particular weight for the feature, and

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the feature and (2) have a label specifying a particular outcome, exceeds a threshold frequency, and, if so, decreasing the current value for the regularization penalty for the feature during the training of the machine learning model.

8. A system for training a machine learning model, the system comprising:

one or more memory devices storing instructions; and

one or more data processing apparatus that are configured to interact with the one or more memory devices, and upon execution of the instructions, perform operations including:

receiving a plurality of training examples, wherein each training example comprises one or more features that identify properties of a respective training instance;

training the machine learning model on the plurality of training examples to determine trained values for weights of the machine learning model, wherein the machine learning model comprises a respective weight for each feature found in any of the training examples processed by the machine learning model, wherein, during the training, the machine learning model processes each of the training examples to determine a predicted output for the training example from current values of the weights for the features in the training example, and wherein training the machine learning model further comprises:

assigning a respective initial value for a regularization penalty for a particular feature, wherein the machine learning model determines the predicted output for each training example that comprises the particular feature by performing computations that include combining a current value for the regularization penalty for the particular feature with the current value of the particular weight for the particular feature, and

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the particular feature and (2) have a label specifying a particular outcome, exceeds a threshold frequency, and, if so, decreasing the current value for the regularization penalty for the particular feature during the training of the machine learning model; and

generating, after completing the training of the machine learning model, a trained machine learning model based on the trained values for the weights whose values have been adjusted based on the respective regularization penalties.

9. The system of claim 8 , wherein training the machine learning model further comprises:

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the particular feature and (2) have a label specifying a particular outcome, does not exceed a threshold frequency, and, if so, increasing the current value for the regularization penalty for the particular feature during the training of the machine learning model.

10. The system of claim 8 , wherein the particular feature belongs to a base template, wherein a base template is a feature category that includes features of a single feature type.

11. The system of claim 8 , wherein the particular feature belongs to a cross-template, wherein a cross-template is a feature category that includes a plurality of features, with each of the plurality of features being of a respective feature type.

12. The system of claim 8 , wherein combining a current value for the regularization penalty for the particular feature with the current value of the particular weight for the particular feature, includes dividing the current value of the particular weight for the particular feature by the current value for the regularization penalty for the particular feature.

13. The system of claim 8 , wherein training the machine learning model further comprises:

for each feature found in any of the training examples:

assigning a respective initial value for a regularization penalty for the feature, wherein the machine learning model determines the predicted output for each training example that comprises the feature by performing computations that include combining a current value for the regularization penalty for the feature with the current value of the particular weight for the feature, and

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the feature and (2) have a label specifying a particular outcome, exceeds a threshold frequency, and, if so, decreasing the current value for the regularization penalty for the feature during the training of the machine learning model.

14. A non-transitory computer readable medium storing instructions that, when executed by one or more data processing apparatus, cause the one or more data processing apparatus to perform operations for training a machine learning model, comprising:

receiving a plurality of training examples, wherein each training example comprises one or more features that identify properties of a respective training instance;

training the machine learning model on the plurality of training examples to determine trained values for weights of the machine learning model, wherein the machine learning model comprises a respective weight for each feature found in any of the training examples processed by the machine learning model, wherein, during the training, the machine learning model processes each of the training examples to determine a predicted output for the training example from current values of the weights for the features in the training example, and wherein training the machine learning model further comprises:

assigning a respective initial value for a regularization penalty for a particular feature, wherein the machine learning model determines the predicted output for each training example that comprises the particular feature by performing computations that include combining a current value for the regularization penalty for the particular feature with the current value of the particular weight for the particular feature, and

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the particular feature and (2) have a label specifying a particular outcome, exceeds a threshold frequency, and, if so, decreasing the current value for the regularization penalty for the particular feature during the training of the machine learning model; and

generating, after completing the training of the machine learning model, a trained machine learning model based on the trained values for the weights whose values have been adjusted based on the respective regularization penalties.

15. The non-transitory computer readable medium of claim 14 , wherein training the machine learning model further comprises:

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the particular feature and (2) have a label specifying a particular outcome, does not exceed a threshold frequency, and, if so, increasing the current value for the regularization penalty for the particular feature during the training of the machine learning model.

16. The non-transitory computer readable medium of claim 14 , wherein the particular feature belongs to a base template, wherein a base template is a feature category that includes features of a single feature type.

17. The non-transitory computer readable medium of claim 14 , wherein the particular feature belongs to a cross-template, wherein a cross-template is a feature category that includes a plurality of features, with each of the plurality of features being of a respective feature type.

18. The non-transitory computer readable medium of claim 14 , wherein combining a current value for the regularization penalty for the particular feature with the current value of the particular weight for the particular feature, includes dividing the current value of the particular weight for the particular feature by the current value for the regularization penalty for the particular feature.

19. The non-transitory computer readable medium of claim 4 , wherein the training examples are potential video viewings, and wherein the predicted output is a prediction of whether a user will view a particular video.

20. The non-transitory computer readable medium of claim 14 , wherein training the machine learning model further comprises:

for each feature found in any of the training examples:

assigning a respective initial value for a regularization penalty for the feature, wherein the machine learning model determines the predicted output for each training example that comprises the feature by performing computations that include combining a current value for the regularization penalty for the feature with the current value of the particular weight for the feature, and

determining whether a frequency of training examples processed by the machine learning model that (1) comprise the feature and (2) have a label specifying a particular outcome, exceeds a threshold frequency, and, if so, decreasing the current value for the regularization penalty for the feature during the training of the machine learning model.