Amazon Technologies, Inc.

Meta Platforms, Inc.

Xiao Zhang

Kevin Christopher Miller

Andrew Bruce Dickinson

David Carl Salyers

Shane Ashley Hall

Christopher Ian Hendrie

Aniket Deepak Divecha

Ralph William Flora

Tobias Lars-Olov Holgers

Logan Michael Gore

Shan Kang

Suthee Chaidaroon

Iman Keivanloo

Jeffrey Thomas Svajlenko

Sidharth Gulati

Shruti Subramaniyam

Yuqing Xing

Anbo Chen

Regina M Joy

Rick Nayar

Systems and methods for training and performing query answering for programming related query answering are disclosed. During a training phase, queries and answers of a training dataset may be encoded as vectors into the Euclidean space. Loss values for query-answer pairs, duplicate queries, and augmented queries may be determined using loss functions and may be used to adjust the model of the encoder. During the operation phase, a query may be received and encoded into the Euclidean space. The encoded answer most proximate to the encoded query may be decoded and provided as the most relevant answer.

Neural search for programming-related query answering

Techniques for forecasting long duration floating holidays in online traffic are described. According to some embodiments, a machine learning service receives a request to train a time series forecast model on time series data of a user, receives an input for the time series forecast model that comprises a first feature weight that represents a first pivot day and a second feature weight that represents a second pivot day, performs a linear interpolation on the first feature weight and the second feature weight for a day between the first pivot day and the second pivot day to generate a linearly interpolated first weight of the first feature weight for a feature vector and a linearly interpolated second weight of the second feature weight for the feature vector, determines a first coefficient for the time series forecast model based at least in part on the time series data of the user, the linearly interpolated first weight of the first feature weight from the feature vector, and the linearly interpolated second weight of the second feature weight from the feature vector, generates, by the time series forecast model comprising the first coefficient, a prediction for a future day, and transmits the prediction to the user.

Forecasting long duration floating holidays in online traffic

An online system receives information describing presentations of a content item to individuals. The online system determines impression frequencies for online system users included among the individuals, de-biases this information, and trains a set of machine-learning models to infer a reach for each of multiple impression frequencies using the de-biased information. The online system predicts the reach for each impression frequency using the models, determines an inferred total number of presentations of the content item based on the inferences, and compares the inferred total number of presentations to a known total number of presentations of the content item. If the inferred total number of presentations is greater than the known total number of presentations, the online system adjusts the inferred reach for a set of impression frequencies by reducing the inferred reach for a highest impression frequency and by increasing the inferred reach for one or more lower impression frequencies.

Inferring a reach of a content item presented to individuals by one or more content publishers for a set of impression frequencies

A control-plane component of a virtual network interface (VNI) multiplexing service assigns one or more VNIs as members of a first interface group. A first VNI of the interface group is attached to a first compute instance. Network traffic directed to a particular endpoint address associated with the first interface group is to be distributed among members of the first interface group by client-side components of the service. The control-plane component propagates membership metadata of the first interface group to the client-side components. In response to a detection of an unhealthy state of the first compute instance, the first VNI is attached to a different compute instance by the control-plane component.

Virtual network interface multiplexing

Virtual  network interface multiplexing

Growing community of inventors

Vienna, VA, United States of America

Xiao Zhang