Abstract: Methods and systems are described herein for risk-adjusted predictive bidding for electronic advertisements. A first computing device receives, from a requesting device loading a webpage, a first request for graphical display source code corresponding to a computerized graphical advertisement display to be inserted into one or more impression opportunities on the webpage. A second computing device coupled to the first computing device transmits a second request to a plurality of third-party computing devices for one or more bids for an impression opportunity of the one or more impression opportunities. The second computing device receives the one or more bids from the third-party computing devices and adjusts at least one of the one or more bids based upon a risk factor received from the third-party computing device that submitted the corresponding bid. The first computing device determines whether to select the impression opportunity based on the adjusted one or more bids.

Abstract: Methods and systems are described herein for risk-adjusted predictive bidding for electronic advertisements. A first computing device receives, from a requesting device loading a webpage, a first request for graphical display source code corresponding to a computerized graphical advertisement display to be inserted into one or more impression opportunities on the webpage. A second computing device coupled to the first computing device transmits a second request to a plurality of third-party computing devices for one or more bids for an impression opportunity of the one or more impression opportunities. The second computing device receives the one or more bids from the third-party computing devices and adjusts at least one of the one or more bids based upon a risk factor received from the third-party computing device that submitted the corresponding bid. The first computing device determines whether to select the impression opportunity based on the adjusted one or more bids.

Abstract: A computer accesses a multidimensional vector space that comprises embeddings of catalog item embeddings, each catalog item having one or more item attributes and corresponding item attribute values, and embeddings of associated item attribute values. An identification of one or more seed catalog items is received. Embeddings of each of one or more seed catalog items and embeddings of associated item attribute values are identified within the vector space. For each one of the seed catalog items, for each of the item attributes of the seed catalog item, an item representation in the vector space is located based upon a combination of the embedding of the seed catalog item and of an embedding of an item attribute value representing an item attribute value that is similar to the corresponding item attribute value of the seed catalog item. For each item representation, one or more embeddings of candidate catalog items that are close to the product representation are selected.