Differential bid generation using machine learning

A system comprising:

• one or more computer processors;
• and
• one or more computer-readable mediums storing instructions that, when executed by the one or more computer processors, cause the system to perform operations comprising:
• accessing historical training data associated a set of candidate users, the historical training data including bid amounts provided to the set of candidate users in previous offers and data indicating whether the previous offers were accepted;
• generating a neural network based on the historical training data, the neural network trained to output an acceptance probability score distribution based on content features that describe an input content item and data that describe an input candidate user, the acceptance probability score distribution indicating likelihoods that the input candidate user will view the input content item in exchange for each bid value from a set of bid values;
• generating a first set of content features that describe a first content item;
• using the neural network, on a first input that is based on the first set of content features, to generate a first respective acceptance probability score distribution for each respective candidate user in the set of candidate users to yield a first set of acceptance probability score distributions, each first respective acceptance probability score distribution indicating likelihoods that the respective candidate user will view the first content item in exchange for each bid value from the set of bid values;
• determining a first subset of the set of candidate users to receive an offer to view the first content item and a first set of bid values corresponding to the first subset of the set of candidate users, the determining of the first subset of candidate users being based on the first set of acceptance probability score distributions, user attributes that describe the set of candidate users, and campaign parameters associated with the first content item;
• and
• providing the offer to view the first content item to at least a first candidate user included in the first subset of the set of candidate users, the offer being based on a first bid value from the first set of bid values, the first bid value corresponding to the first candidate user.

A non-transitory computer-readable medium storing instructions that, when executed by one or more computer processors of one or more computing devices, cause the one or more computing devices to perform operations comprising:

• accessing historical training data associated a set of candidate users, the historical training data including bid amounts provided to the set of candidate users in previous offers and data indicating whether the previous offers were accepted;
• generating a neural network based on the historical training data, the neural network trained to output an acceptance probability score distribution based on content features that describe an input content item and data that describe an input candidate user, the acceptance probability score distribution indicating likelihoods that the input candidate user will view the input content item in exchange for each bid value from a set of bid values;
• generating a first set of content features that describe a first content item;
• using the neural network, on a first input that is based on the first set of content features, to generate a first respective acceptance probability score distribution for each respective candidate user in the set of candidate users to yield a first set of acceptance probability score distributions, each first respective acceptance probability score distribution indicating likelihoods that the respective candidate user will view the first content item in exchange for each bid value from the set of bid values;
• determining a first subset of the set of candidate users to receive an offer to view the first content item and a first set of bid values corresponding to the first subset of the set of candidate users, the determining of the first subset of candidate users being based on the first set of acceptance probability score distributions, user attributes that describe the set of candidate users, and campaign parameters associated with the first content item;
• and
• providing the offer to view the first content item to at least a first candidate user included in the first subset of the set of candidate users, the offer being based on a first bid value from the first set of bid values, the first bid value corresponding to the first candidate user.