Abstract: A computer-implemented method for layered training of machine-learning architectures includes receiving a plurality of data elements wherein each data element is associated with a timestamp, determining a training window for each model layer of a layered stack of model layers, determining a plurality of training data elements for each training window by identifying the data elements with timestamps corresponding to each of the training windows, identifying a previous checkpoint for each model layer wherein the previous checkpoint for each model layer is generated by a parent model layer, training each model layer with the determined training data elements for each model layer and the identified previous checkpoint for each model layer, generating a plurality of current checkpoints wherein each current checkpoint of the plurality of current checkpoints is associated with a model layer, and storing the plurality of current checkpoints at the memory.

Abstract: A computer-implemented method for layered training of machine-learning architectures includes receiving a plurality of data elements wherein each data element is associated with a timestamp, determining a training window for each model layer of a layered stack of model layers, determining a plurality of training data elements for each training window by identifying the data elements with timestamps corresponding to each of the training windows, identifying a previous checkpoint for each model layer wherein the previous checkpoint for each model layer is generated by a parent model layer, training each model layer with the determined training data elements for each model layer and the identified previous checkpoint for each model layer, generating a plurality of current checkpoints wherein each current checkpoint of the plurality of current checkpoints is associated with a model layer, and storing the plurality of current checkpoints at the memory.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for classifying objects. Object (e.g., Website) classification can be performed using multiple models that are implemented in a distributed processing environment. In response to receiving a request to classify an object based on attribute values that characterize the object models that are configured to classify objects based on the attribute values are identified and the attribute values are provided to the identified models. Based on classification data received from selected models is used to generate a final classification for the object. The final classification can specify or correspond to a presentation availability status of the object that can be used to control availability of the object. For example, a Website having a final classification of “blacklisted” can be made unavailable for presentation based on the classification result.