Abstract: Briefly, embodiments of methods and/or systems of generating preference indices for contiguous portions of digital images are disclosed. For one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. The developed parameters may be transferred to a neural network utilized to generate signal sample value levels corresponding to preference indices for contiguous portions of digital images.

Abstract: The present teaching relates to recommending content items to a user based on tensor factorization. In one example, a request is received for recommending content items to the user. Tensor data related to a plurality of users and a plurality of content items are obtained based on the request. The tensor data is decomposed into a plurality of sub-tensors based on a prior probability distribution. At least one bound is determined for a tensor factorization model that is generated based on the prior probability distribution. One or more items interesting to the user are predicted based on the at least one bound and the plurality of sub-tensors. At least one of the one or more items is recommended to the user as a response to the request.

Abstract: Users may consume and/or share information through various types of content items. For example, user may post a family photo through a social network, create a running blog through a microblogging service, etc. Because users may be overwhelmed by the amount of available content items, it may be advantageous to recommend content items, such as blogs to follow, to users. Accordingly, inductive matrix completion is used to evaluate user interactions with content items (e.g., a user following a blog), content item features (e.g., text and/or images of a blog is evaluated to identify a topic of the blog), and/or user features (e.g., a user liking or reblogging a blog, user demographics, user interests, etc.) to determine whether to recommend a content item to a user. Additionally, graph proximity is used to recommend content items based upon weights of edges connecting user nodes to content item nodes within a directed graph.

Abstract: The technologies described herein serve contextually relevant advertisements under a guaranteed advertisement campaign. A publisher retrieves a guaranteed advertisement campaign related to a webpage available for serving an advertisement, and identifies a set of advertisements relating to the guaranteed advertisement campaign. Advertisement selecting circuitry of the publisher determines whether an advertisement that is contextually relevant to content published at the webpage is present in the set of advertisements. If there is no contextually relevant advertisement in the set of advertisements, the advertisement selecting circuitry selects an alternative advertisement from the set of advertisements that minimizes an under-delivery risk related to the guaranteed advertisement campaign. If there is a contextually relevant advertisement in the set of advertisements, the advertisement selecting circuitry selects the contextually relevant advertisement.

Abstract: Systems, methods, and apparatuses are disclosed for generating mapping data structures based on predicted relationships across tracking data obtained from tracking online browsing histories of users to a network of websites.

Abstract: The present teaching relates to dynamically generate a score floor model that provides a dynamically determined threshold metric to be used to select future bids that satisfy a first and second set of metrics. Feedback information is first received that relates to a plurality of bids previously submitted by a bidder to a publisher, wherein the feedback information indicates whether each of the plurality of bids has led to an impression. The first set of metrics specified by the publisher is obtained and the second set of metrics is retrieved. The score floor model with respect to the publisher is dynamically updated based on the feedback information as well as the first and second sets of metrics.

Abstract: Systems, devices, and methods are disclosed for predicting a dynamic floor price for increasing cleared revenue cleared after a winning bid is determined in an online bid auction. The dynamic floor price is predicted from a cascading classifier strategy implemented through a series of cascading machine learning based classifier models that have been trained.

Abstract: Users may consume and/or share information through various types of content items. For example, user may post a family photo through a social network, create a running blog through a microblogging service, etc. Because users may be overwhelmed by the amount of available content items, it may be advantageous to recommend content items, such as blogs to follow, to users. Accordingly, inductive matrix completion is used to evaluate user interactions with content items (e.g., a user following a blog), content item features (e.g., text and/or images of a blog is evaluated to identify a topic of the blog), and/or user features (e.g., a user liking or reblogging a blog, user demographics, user interests, etc.) to determine whether to recommend a content item to a user. Additionally, graph proximity is used to recommend content items based upon weights of edges connecting user nodes to content item nodes within a directed graph.

Abstract: Users may consume and/or share information through various types of content items. For example, user may post a family photo through a social network, create a running blog through a microblogging service, etc. Because users may be overwhelmed by the amount of available content items, it may be advantageous to recommend content items, such as blogs to follow, to users. Accordingly, inductive matrix completion is used to evaluate user interactions with content items (e.g., a user following a blog), content item features (e.g., text and/or images of a blog is evaluated to identify a topic of the blog), and/or user features (e.g., a user liking or reblogging a blog, user demographics, user interests, etc.) to determine whether to recommend a content item to a user. Additionally, graph proximity is used to recommend content items based upon weights of edges connecting user nodes to content item nodes within a directed graph.

Abstract: To maximize the accuracy and efficiency of predicting users that will enjoy targeted content, a proposed content selection solution looks to combine a first strategy of utilizing selection rules with a second strategy of utilizing machine based learning models. By combining the selection rules-based approach and the machine learning model-based approach, the proposed content selection solution is able to consider and recommend a wider range of users for each available content.

Abstract: Systems, devices, and methods are disclosed for predicting a dynamic floor price for increasing cleared revenue cleared after a winning bid is determined in an online bid auction. The dynamic floor price is predicted from a cascading classifier strategy implemented through a series of cascading machine learning based classifier models that have been trained.

Abstract: The present teaching relates to dynamically generate a score floor model that provides a dynamically determined threshold metric to be used to select future bids that satisfy a first and second set of metrics. Feedback information is first received that relates to a plurality of bids previously submitted by a bidder to a publisher, wherein the feedback information indicates whether each of the plurality of bids has led to an impression. The first set of metrics specified by the publisher is obtained and the second set of metrics is retrieved. The score floor model with respect to the publisher is dynamically updated based on the feedback information as well as the first and second sets of metrics.

Abstract: The present teaching relates to recommending content items to a user based on tensor factorization. In one example, a request is received for recommending content items to the user. Tensor data related to a plurality of users and a plurality of content items are obtained based on the request. The tensor data is decomposed into a plurality of sub-tensors based on a prior probability distribution. At least one bound is determined for a tensor factorization model that is generated based on the prior probability distribution. One or more items interesting to the user are predicted based on the at least one bound and the plurality of sub-tensors. At least one of the one or more items is recommended to the user as a response to the request.

Abstract: To maximize the accuracy and efficiency of predicting users that will enjoy targeted content, a proposed content selection solution looks to combine a first strategy of utilizing selection rules with a second strategy of utilizing machine based learning models. By combining the selection rules-based approach and the machine learning model-based approach, the proposed content selection solution is able to consider and recommend a wider range of users for each available content.

Abstract: Described herein are techniques and systems for online ad campaign pacing. The techniques described herein use budget allocation along with the estimations of bids and response rates. With use of budget allocation, the techniques can use budget pacing to enhance impressions and maximize desired responses, such as desired click-through rates. These techniques focus on enhancing pacing and performance of ad campaigns, such as enhancing performance across distinct and/or unified online ad marketplaces. These techniques are especially useful in the context of a demand-side platform (DSP). In some examples, the techniques assume that impression supply is much larger than advertiser demand for impressions of their ads, so such techniques focus on selecting high performing inventory of ad space. Yet, with such a focus, a smooth or consistent delivery of ads over time is used.

Abstract: Embodiments of the invention present an approach to conversion rate estimation which relies on using past performance observations along user, publisher, and advertiser data hierarchies. More specifically, embodiments of the invention model the conversion event at different select hierarchical levels with separate binomial distributions and estimate the distribution parameters individually. It is shown how to combine these individual estimators using logistic regression to identify conversion events accurately. Embodiments of the invention can also handle many practical issues, such as data imbalance, missing data, and output probability calibration, which render this estimation problem more difficult for a real-world implementation of the approach.

Abstract: Systems, methods, and apparatuses are disclosed for generating mapping data structures based on predicted relationships across tracking data obtained from tracking online browsing histories of users to a network of websites.

Abstract: Briefly, embodiments of methods and/or systems of generating preference indices for contiguous portions of digital images are disclosed. For one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. The developed parameters may be transferred to a neural network utilized to generate signal sample value levels corresponding to preference indices for contiguous portions of digital images.

Abstract: Embodiments described herein facilitate or enhance the implementation of image recognition processes which can perform recognition on images to identify objects and/or faces by class or by people.

Abstract: Embodiments described herein facilitate or enhance the implementation of image recognition processes which can perform recognition on images to identify objects and/or faces by class or by people.