Abstract: Systems and techniques are provided for template exploration in a large-scale machine learning system. A method may include obtaining multiple base templates, each base template comprising multiple features. A template performance score may be obtained for each base template and a first base template may be selected from among the multiple base templates based on the template performance score of the first base template. Multiple cross-templates may be constructed by generating a cross-template of the selected first base template and each of the multiple base templates. Performance of a machine learning model may be tested based on each cross-template to generate a cross-template performance score for each of the cross-templates. A first cross-template may be selected from among the multiple cross-templates based on the cross-template performance score of the cross-template. Accordingly, the first cross-template may be added to the machine learning model.

Abstract: Systems and techniques are provided for template exploration in a large-scale machine learning system. A method may include obtaining multiple base templates, each base template comprising multiple features. A template performance score may be obtained for each base template and a first base template may be selected from among the multiple base templates based on the template performance score of the first base template. Multiple cross-templates may be constructed by generating a cross-template of the selected first base template and each of the multiple base templates. Performance of a machine learning model may be tested based on each cross-template to generate a cross-template performance score for each of the cross-templates. A first cross-template may be selected from among the multiple cross-templates based on the cross-template performance score of the cross-template. Accordingly, the first cross-template may be added to the machine learning model.

Abstract: A method includes presenting a list of one or more videos via a user interface, receiving a selection of a target video to watch from the list, playing the target video in the user interface, and updating the user interface to present one or more suggested videos concurrently with playback of the target video. The one or more suggested videos are predicted to be watched by a user for at least a threshold duration.

Abstract: A system receives a user request for a media item and identifies candidate media items for suggesting to the user with the media item. The system predicts a user consumption time for each candidate media item and selects a sub-set of the candidate media items that have higher predicted user consumption times. The system provides the requested media item with the sub-set of the candidate media items.

Abstract: Mechanisms for recommending content items based on topics are provided. In some implementations, a method for recommending content items is provided that includes: determining a plurality of accessed content items associated with a user, wherein each of the plurality of content items is associated with a plurality of topics; determining the plurality of topics associated with each of the plurality of accessed content items; generating a model of user interests based on the plurality of topics, wherein the model implements a machine learning technique to determine a plurality of weights for assigning to each of the plurality of topics; applying the model to determine, for a plurality of content items, a probability that the user would watch a content item of the plurality of content items; ranking the plurality of content items based on the determined probabilities; and selecting a subset of the plurality of content items to recommend to the user based on the ranked content items.

Abstract: Mechanisms for recommending content items based on topics are provided. In some implementations, a method for recommending content items is provided that includes: determining a plurality of accessed content items associated with a user, wherein each of the plurality of content items is associated with a plurality of topics; determining the plurality of topics associated with each of the plurality of accessed content items; generating a model of user interests based on the plurality of topics, wherein the model implements a machine learning technique to determine a plurality of weights for assigning to each of the plurality of topics; applying the model to determine, for a plurality of content items, a probability that the user would watch a content item of the plurality of content items; ranking the plurality of content items based on the determined probabilities; and selecting a subset of the plurality of content items to recommend to the user based on the ranked content items.

Abstract: Systems and techniques are disclosed for training a machine learning model based on one or more regularization penalties associated with one or more features. A template having a lower regularization penalty may be given preference over a template having a higher regularization penalty. A regularization penalty may be determined based on domain knowledge. A restrictive regularization penalty may be assigned to a template based on determining that a template occurrence is below a stability threshold and may be modified if the template occurrence meets or exceeds the stability threshold.

Abstract: Mechanisms for recommending content items based on topics are provided. In some implementations, a method for recommending content items is provided that includes: determining a plurality of accessed content items associated with a user, wherein each of the plurality of content items is associated with a plurality of topics; determining the plurality of topics associated with each of the plurality of accessed content items; generating a model of user interests based on the plurality of topics, wherein the model implements a machine learning technique to determine a plurality of weights for assigning to each of the plurality of topics; applying the model to determine, for a plurality of content items, a probability that the user would watch a content item of the plurality of content items; ranking the plurality of content items based on the determined probabilities; and selecting a subset of the plurality of content items to recommend to the user based on the ranked content items.

Abstract: Systems and techniques are disclosed for training a machine learning model based on one or more regularization penalties associated with one or more features. A template having a lower regularization penalty may be given preference over a template having a higher regularization penalty. A regularization penalty may be determined based on domain knowledge. A restrictive regularization penalty may be assigned to a template based on determining that a template occurrence is below a stability threshold and may be modified if the template occurrence meets or exceeds the stability threshold.