Modern web-scale behavioral targeting platforms leverage historical activity of billions of users to predict user interests and inclinations, and consequently future activities. Future activities of particular interest involve purchases or transactions, and are referred to as conversions. Unlike ad-clicks, conversions directly translate to advertiser's revenue, and thus provide a very concrete metric for return on advertising investment. A typical behavioral targeting system faces two main challenges: the web-scale amounts of user histories to process on a daily basis, and the relative sparsity of conversions (compared to clicks in a traditional setting). These challenges call for generation of effective and efficient user profiles. Most existing works use the historical intensity of a user's interest in various topics to model future interest. In this paper we explore how the change in user behavior can be used to predict future actions and show how it complements the traditional models of decaying interest and action recency to build a complete picture about the user interests and better predict conversions. Our evaluation over a real-world set of campaigns indicates that the combination of change of interest, decaying intensity, and action recency helps in: 1) scoring significant improvements in optimizing for conversions over traditional baselines, 2) substantially improving the targeting efficiency for campaigns with highly sparse conversions, and 3) highly reducing the overall history sizes used in targeting. Furthermore, our techniques have been deployed to production and scored a substantial improvement in targeting performance while imposing a negligible overhead in terms of overall platform running time.

A typical behavioral targeting system optimizing purchase activities, called conversions, faces two main challenges: the web-scale amounts of user histories to process on a daily basis, and the relative sparsity of conversions. In this paper, we try to address these challenges through feature selection. We formulate a multi-task (or group) feature-selection problem among a set of related tasks (sharing a common set of features), namely advertising campaigns. We apply a group-sparse penalty consisting of a combination of an l1 and l2 penalty and an associated fast optimization algorithm for distributed parameter estimation. Our algorithm relies on a variant of the well known Fast Iterative Thresholding Algorithm (FISTA), a closed-form solution for mixed norm programming and a distributed subgradient oracle. To efficiently handle web-scale user histories, we present a distributed inference algorithm for the problem that scales to billions of instances and millions of attributes. We show the superiority of our algorithm in terms of both sparsity and ROC performance over baseline feature selection methods (both single-task -regularization and multi-task mutual-information gain).

We present a novel algorithm that takes as input an uncalibrated unordered set of spherical panoramic images and outputs their relative pose up to a global scale. The panoramas contain both indoor and outdoor shots and each set was taken in a particular indoor location e.g. a bakery or a restaurant. The estimated pose is used to build a map of the location, and allow easy visual navigation and exploration in the spirit of Google's Street View. We also present a dataset of 9 sets of panoramas, together with an annotation tool and ground truth point correspondences. The manual annotations were used to obtain ground truth relative pose, and to quantitatively evaluate the different parameters of our algorithm, and can be used to benchmark different approaches. We show excellent results on the dataset and point out future work.

Latent variable techniques are pivotal in tasks ranging from predicting user click patterns and targeting ads to organizing the news and managing user generated content. Latent variable techniques like topic modeling, clustering, and subspace estimation provide substantial insight into the latent structure of complex data with little or no external guidance making them ideal for reasoning about large-scale, rapidly evolving datasets. Unfortunately, due to the data dependencies and global state introduced by latent variables and the iterative nature of latent variable inference, latent-variable techniques are often prohibitively expensive to apply to large-scale, streaming datasets.

Understanding what interests and delights users is critical to effective behavioral targeting, especially in information-poor contexts. As users interact with content and advertising, their passive behavior can reveal their interests towards advertising. Two issues are critical for building effective targeting methods: what metric to optimize for and how to optimize. More specifically, we first attempt to understand what the learning objective should be for behavioral targeting so as to maximize advertiser's performance. While most popular advertising methods optimize for user clicks, as we will show, maximizing clicks does not necessarily imply maximizing purchase activities or transactions, called conversions, which directly translate to advertiser's revenue. In this work we focus on conversions which makes a more relevant metric but also the more challenging one. Second is the issue of how to represent and combine the plethora of user activities such as search queries, page views, ad clicks to perform the targeting. We investigate several sources of user activities as well as methods for inferring conversion likelihood given the activities. We also explore the role played by the temporal aspect of user activities for targeting, e.g., how recent activities compare to the old ones. Based on a rigorous offline empirical evaluation over 200 individual advertising campaigns, we arrive at what we believe are best practices for behavioral targeting. We deploy our approach over live user traffic to demonstrate its superiority over existing state-of-the-art targeting methods.

Historical user activity is key for building user profiles to predict the user behavior and affinities in many web applications such as targeting of online advertising, content personalization and social recommendations. User profiles are temporal, and changes in a user's activity patterns are particularly useful for improved prediction and recommendation. For instance, an increased interest in car-related web pages may well suggest that the user might be shopping for a new vehicle.In this paper we present a comprehensive statistical framework for user profiling based on topic models which is able to capture such effects in a fully \emph{unsupervised} fashion. Our method models topical interests of a user dynamically where both the user association with the topics and the topics themselves are allowed to vary over time, thus ensuring that the profiles remain current.

Sensor networks will shortly consist of globally deployed sensors providing real-time geo-centric information to users. Particularly, users with mobile devices will issue ad-hoc queries usually from within, or nearby, the queried area. In this paper, we propose an in- network data-centric storage (DCS) scheme, namely the spatio-temporal data-centric storage (STDCS) scheme, to efficiently answer these mobile user queries. STDCS is designed to maintain load-balancing among sensors to cope with query hotspots in the network. It is different from previous proposals in two aspects. First, it is based on the novel idea of using a temporally evolving spatial indexing scheme to balance querying load among sensors. Furthermore, STDCS uses dynamic mechanisms for query hotspot detection and decomposition. We conducted extensive simulations that showed our scheme's superiority to both local storage and spatial indexing (the only known geo-centric storage schemes) whenever experiencing query hotspots of different sizes.

The rapid growth of the number of videos in YouTube provides enormous potential for users to find content of interest to them. Unfortunately, given the difficulty of searching videos, the size of the video repository also makes the discovery of new content a daunting task. In this paper, we present a novel method based upon the analysis of the entire user-video graph to provide personalized video suggestions for users. The resulting algorithm, termed Adsorption, provides a simple method to efficiently propagate preference information through a variety of graphs. We extensively test the results of the recommendations on a three month snapshot of live data from YouTube.

The expected architecture of next-generation sensornets raises the need for load-balanced point-to-point routing protocols to cope with different sources of traffic skewness. Query hotspots are one of these skewness sources that highly impact the sensornet performance. In this paper, we present a set of content-based load-balancing primitives to be used on top of any point- to-point routing protocol in order to detect and decompose query hotspots. Our schemes are based on local hotspot detection by the sensors targeted by queries. Hotspots are then decomposed by avoiding duplication when forwarding results to the query issuers. Our simulation results show the high benefit, in terms of network lifetime and throughput, of using our schemes to load-balance query hotspots of different sizes.

We propose an In-Network Data-Centric Storage (INDCS) scheme for answering ad-hoc queries in sensor networks. Previously proposed In-Network Storage (INS) schemes suffered from Storage Hot-Spots that are formed if either the sensors' locations are not uniformly distributed over the coverage area, or the distribution of sensor readings is not uniform over the range of possible reading values. Our K-D tree based Data-Centric Storage (KDDCS) scheme maintains the invariant that the storage of events is distributed reasonably uniformly among the sensors. KDDCS is composed of a set of distributed algorithms whose running time is within a poly-log factor of the diameter of the network. The number of messages any sensor has to send, as well as the bits in those messages, is poly-logarithmic in the number of sensors. Load balancing in KDDCS is based on defining and distributively solving a theoretical problem that we call the Weighted Split Median problem. In addition to analytical bounds on KDDCS individual algorithms, we provide experimental evidence of our scheme's general efficiency, as well as its ability to avoid the formation of storage hot-spots of various sizes, unlike all previous INDCS schemes.

Arising when a large percentage of queries is accessing data stored in few sensor nodes, query hot-spots reduce the quality of data (QoD) and the lifetime of the sensor network. All current in-network data-centric storage (IN-DCS) schemes fail to deal with query hot-spots resulting from skewed query loads as well as skewed sensor deployments. In this paper, we present two algorithms to locally detect and decompose query hot-spots, namely zone partitioning (ZP) and zone partial replication (ZPR). We build both algorithms on top of the DIM scheme, which has been shown to exhibit the best performance among all INDCS schemes. Experimental evaluation illustrates the efficiency of ZP/ZPR in decomposing query hot-spots while increasing QoD as well as energy savings by balancing energy consumption among sensor nodes