International Business Machines Corporation

Google Inc.

Amazon Technologies, Inc.

Sheng Ma

Genady Grabarnik

Joseph L Hellerstein

Chang-Shing Perng

Laura Grace Book

James Robert Koehler

Irina Rish

Evgeny Skvortsov

Jiayu Peng

David Hutchinson Thoenen

Xichen Huang

Abdolreza Salahshour

Benjamin Kreuter

Craig Wright

Mark Alan Brodie

Arthur Asuncion

Natalia V Odintsova

Fan Zhang

Scott D Schneider

Alina Beygelzimer

Joseph Sean Cahill Goodknight Knightbrook

Michael Daub

Chang-shing Perng

Moon Jung Kim

Lev Kozakov

Balan Subramanian

David Alan Rabenhorst

Sarah Masters

Philip Shi-lung Yu

Jonathan Lenchner

Shang Qing Guo

Yurdaer Nezihi Doganata

Youssef Drissi

Haixun Wang

Lu Zhang

Tong-Haing Fin

Juan Leon Rodriguez

Paula Marie Despins

Luanne M Burns

Thomas A Vaughan

Hancheng Ge

Jia Chen

Zachary Charles Frazier

Steve Lin

Feng Liang

Steven S Demion

Benjamin R Kreuter

Systems and methods for generating min-increment counting bloom filters to determine count and frequency of device identifiers and attributes in a networking environment are disclosed. The system can maintain a set of data records including device identifiers and attributes associated with device in a network. The system can generate a vector comprising coordinates corresponding to counter registers. The system can identify hash functions to update a counting bloom filter. The system can hash the data records to extract index values pointing to a set of counter registers. The system can increment the positions in the min-increment counting bloom filter corresponding to the minimum values of the counter registers. The system can obtain an aggregated public key comprising a public key. The system can encrypt the counter registers using the aggregated shared key to generate an encrypted vector. The system can transmit the encrypted vector to a networked worker computing device.

Secure multi-party reach and frequency estimation

Systems and methods are disclosed herein for improved per-frequency counting systems that record interactions between individuals and a group of providers while maintaining differential privacy. A protocol may be defined that specifies frequency bins to categorize identifiers corresponding to individuals. A provider may generate a plurality of private sketches, each corresponding to a plurality of frequencies defined in the protocol. Frequency data is determined for each identifier. Identifiers are encoded into the private sketches corresponding to the identifiers' associated frequency. The plurality of private sketches from each provider in the group of providers are combined to generate a deduplicated distribution across the group. In one implementation, the private sketches of each provider are sequentially merged until all sketches have been combined, from which the total distribution can be estimated.

Differentially private frequency deduplication

At least one aspect is directed to determining an estimate of an intersection of user identifiers in a first set of user identifiers and a second set of user identifiers. The first and second sets of user identifiers can be populated with user identifiers that have interacted with the same content item or content item campaign. Estimates of intersections of the first and the second sets can be determined based on a binomial vector approach, a vector of counts approach, or a hybrid approach. The binomial vector approach generates vectors based on k hashes of each user identifier in the first set and summing the vectors to generate a first vector. The intersection can be determined based on a dot product of the first vector and a second vector similarly generated from the second set of user identifiers.

Systems and methods for privacy preserving determination of intersections of sets of user identifiers

Systems, methods, and computer-readable media are disclosed for modeled advertisement conversion attributions. An example method may include receiving first input data comprising first advertisement impression data and first advertisement conversion data, wherein the first input data includes one or more user identifiers associated with both the advertisement impression data and advertisement conversion data. The example method may also include training one or more machine learning models using the first input data. The example method may also include receiving second input data comprising second advertisement impression data, wherein user identifiers are unavailable for the second input data. The example method may also include determining, using the one or more machine learning models, second predicted conversion data associated with the second input data.

Systems and methods for modeled advertisement conversion attributions

Methods, systems, and apparatus include computer programs encoded on a computer-readable storage medium, including a method for discovering unique entities over multiple devices. A virtual pool of entities is created and divided into subpools, each including fewer than all entities. Subpools are subdivided into delta pools. Cookies are recorded for each delta pool when the particular portion of content is presented to or accessed by entities in the delta pool. Recorded cookies are divided into cookie types based on cookie characteristics. Machine learning and statistical analysis algorithms are used to automatically determine sizes of delta pools and probabilities of each cookie type being classified as belonging to particular delta pools. Virtual entities are assigned from the virtual pool to each of the recorded cookies that were recorded when the particular portion of content was presented. A number of unique entities that accessed the particular portion of content is determined.

Discovery of unique entities across multiple devices

Methods, systems, and apparatus, including computer program products, for processing events related to presented content. In one aspect, a method includes determining a time window count of a number of advertising events associated with an advertisement during at least one time window before a conversion event; and determining a credit that represents a strength of an association between the advertisement and the conversion event, wherein determining a credit includes selecting a weighting model for the at least one time window count.

Conversion crediting

A framework for content delivery includes a content distribution manager operable to select content for distribution in response to a request for content. The selection of content may be made based in part on scheduling information associated with sources of available content and selected by content providers. The selection may further be based in part on a content selection policy selected by publishers who initiate requests for content.

Distributing content

Improved problem diagnosis techniques for use in accordance with computing systems, e.g., distributed computing systems, are disclosed. In one aspect of the invention, a technique for diagnosing a problem associated with a computing system comprises the following steps/operations. One or more probes are executed in accordance with at least a portion of a previously selected probe schedule. When a result of one or more of the probes of the previously selected probe schedule indicates, at least, a potential problem associated with the computing system, one or more probes which optimize at least one criterion are selected in real-time. The one or more selected probes are executed so as to diagnose the potential problem.

Active probing for real-time diagnosis

Growing community of inventors

Briarcliff Manor, NY, United States of America

Sheng Ma