Google Inc.

Thomson Licensing

Wenjie Jiang

Azalia Mirhoseini

James Laudon

Shen Wang

Jeffrey Adgate Dean

Quoc V Le

Chian-Min Richard Ho

Anna Darling Goldie

Mustafa Nazim Yazgan

Roger David Carpenter

Young-Joon Lee

Ebrahim Songhori

Omkar Pathak

Kavya Srinivasa Setty

Anurag Sharma

Matthew Johnston

Rodolfo Enrique Alvizu Gomez

Min Wen

Bikash Koley

Emilie Jeanne Anne Danna

Xiaoxue Zhao

Jon Vaver

Stefan F Schnabl

Laurent Massoulie

Ajay Kumar Bangla

John Huang

Alireza Ghaffarkhah

Efstratios Ioannidis

Christoph Albrecht

Fabio Picconi

Arjun Satyapal

Benjamin Preskill

Aspects of the disclosure are directed to automatically determining floor planning in chips, which factors in memory macro alignment. A deep reinforcement learning (RL) agent can be trained to determine optimal placements for the memory macros, where memory macro alignment can be included as a regularization cost to be added to the placement objective as a RL reward. Tradeoffs between the placement objective and alignment of macros can be controlled by a tunable alignment parameter.

Alignment cost for integrated circuit placement

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a computer chip placement. One of the methods includes obtaining netlist data for a computer chip; and generating a computer chip placement, comprising placing a respective macro node at each time step in a sequence comprising a plurality of time steps, the placing comprising, for each time step: generating an input representation for the time step; processing the input representation using a node placement neural network having a plurality of network parameters, wherein the node placement neural network is configured to process the input representation in accordance with current values of the network parameters to generate a score distribution over a plurality of positions on the surface of the computer chip; and assigning the macro node to be placed at the time step to a position from the plurality of positions using the score distribution.

Generating integrated circuit placements using neural networks

Methods, systems, and apparatus, including computer-readable storage media, optimizing interior gateway protocol (IGP) metrics using reinforcement learning (RL) for a network domain. The system can receive a topology (G) of a network domain, a set of flows (F), and an objective function. The system can optimize, using reinforcement learning, the objective function based on the received topology and the one or more flows F. The system can determine updated IGP metrics based on the optimization of the objective function. The IGP metrics for the metric domain may be updated with the updated IGP metrics.

Assigning routing paths based on interior gateway protocol metric optimization

A content distribution network, including: a content server that stores a set of content items; a plurality of gateways for storing and serving content requests to a subset of the content items, the plurality of gateways being grouped into a plurality of classes of gateways; and a plurality of class trackers corresponding to the plurality of classes; wherein each class tracker manages a placement of content items and an assignment of content requests for its class of gateways; wherein the plurality of class trackers exchange congestion signals among themselves; wherein, for each content item, each class tracker determines a fraction of gateways in its class of gateways that store the content item and a rate of content requests that are forwarded to the content server or other class trackers, based on requests for the content item entering its class and the congestion signals received from other class trackers.

Method and apparatus for transferring content among large clusters of storage devices to achieve a target replication distribution

The present disclosure provides a probabilistic framework that can calculate the probability of fulfilling demands for a given set of traffic flows. In some implementations, the probability of fulfilling demands can be based on the probability of infrastructure component failures, shared risk link groups derived from a cross-layer network topology, and traffic engineering (TE) considerations. The consideration of the cross-layer network topology enables the systems and methods described herein to account for the relationship between the physical and logical topologies.

Parallelized network traffic flow availability simulation using stochastic process and traffic engineering algorithms

The present disclosure includes systems and techniques relating to identifying value marketing activities. In some implementations, an apparatus, systems, or methods can include receiving conversion path information including data relating to user interactions with a content item associated with a marketing activity, determining a first attribution credit by applying a first attribution model to the received information, and a second attribution credit by applying a second attribution model to the received information, determining an attribution contrast ratio based on the first and second attribution credit, identifying an opportunity based on the determined attribution contrast ratio, and presenting a recommendation for the marketing activity based on the identified opportunity.

Attribution marketing recommendations

Growing community of inventors

Mountain View, CA, United States of America

Wenjie Jiang