Amazon Technologies, Inc.

Balakrishnan Narayanaswamy

Yannis Papakonstantinou

Vuk Ercegovac

Gaurav Saxena

Rashmi Gangadharaiah

Charles Elkan

Gokul Soundararajan

Sriram Krishnamurthy

George Constantin Caragea

Mohammad Rezaur Rahman

Enrico Siragusa

Nikolaos Koulouris

Jiayuan Chen

Ippokratis Pandis

Naresh Chainani

Jeremiah Wilton

Christos Faloutsos

Vijay Gopal Joshi

Fabian Oliver Nagel

Maxym Kharchenko

Andre Hernich

Zhao Song

Panagiotis Parchas

Tim Kraska

Davide Pagano

Kari E J Torkkola

Zhihao Cen

Mario Guerriero

Joshua Tobey Oberwetter

Kyle Henderson Hailey

Kaihui Zheng

Zhengchun Liu

Ru He

Wen-Yu Hua

Alexander Matthew Lamb

Vikramank Yogendra Singh

A future workload may be predicted for a database system using analysis of queries submitted for execution. A feature vector for a query may be determined according to a query plan for the query. If the feature vector has not been previously seen, or has not been sufficiently seen, by the database system, a machine learning inference may be used to predict performance characteristics of the query, the machine learning system trained using previous feature vectors and performance characteristics of executed queries. If the feature vector has been sufficiently seen previously by the database system, a history of performance characteristics of previous queries with similar or the same feature vector may be used to predict the performance characteristics. The predictions may then be used to configure or reconfigure processing cluster(s) of the database system.

Query performance prediction using multiple experts

An anomalous period of operation of a database management system is detected by analyzing a time series of data points indicating the number of database queries pending processing by the system. Conditions associated with execution of the pending database queries are recorded and analyzed to identify conditions correlated with the anomalous period of operation. A recommendation for tuning the database is generated based on analysis of the conditions.

Assisted database anomaly mitigation

Queries that reference materialized views may be rewritten to compensate for stale materialized views. A query may be received that references a materialized view. A determination may be made as to whether the materialized view is stale. For a stale materialized view, the query may be rewritten to generate a query plan that obtains changes made to base tables for the materialized view not included in the stale materialized view and considers the change(s) when generating a result for the query from the materialized view. The rewritten query may then be performed to provide a result as if the materialized view were up-to-date.

Rewriting queries to compensate for stale materialized views

In response to determining that a particular sequence of natural language input has been generated by a first entity participating in a multi-interaction dialog, a first representation of accumulated dialog state associated with the sequence is obtained from a machine learning model at an artificial intelligence service. Based on the first representation, a state response entry is selected from a collection of state response entries. The state response entry indicates a mapping between a second representation of accumulated dialog state, and a response recorded in a training example of the model. The recorded response is implemented.

Artificial intelligence system using hybrid technique for task-oriented dialog management

Scaling of query processing resources for efficient utilization and performance is implemented for a database service. A query is received via a network endpoint associated with a database managed by a database service. Respective response times predicted for the query using different query processing configurations available to perform the query are determined. Those query processing configurations with response times that exceed a variability threshold determined for the query may be excluded. A remaining query processing configuration may then be selected to perform the query.

Scaling query processing resources for efficient utilization and performance

A database system may use a machine learning model creation system to create a machine learning model from data stored in the database system responsive to a request from a client. The database system may obtain an executable version of the machine learning model, based on an uncompiled hardware agnostic version of the machine learning model, according to the hardware configuration of one or more computing resources selected by the database system to perform requests to the database system that invoke the machine learning model to generate predictions.

Dynamic compilation of machine learning models based on hardware configurations

A database system may include a machine learning model which may be used to perform various data analytics for data stored in the database system. In response to a request to invoke the machine learning model to generate a prediction from data stored in the database system, the database system may perform one or more optimization operations, as part of a query plan, to prepare the data to make it suitable for use by the machine learning model.

Integrating query optimization with machine learning model prediction

Materialized views for a database system may be automatically refreshed according to performance benefits. Materialized views may be ordered according to determined performance benefits for the materialized views indicating the performance benefit obtained when a materialized view is used to perform a query at the database system. Materialized views may be selected for refresh operations according to the ordering based on a capacity of the database system to perform refresh operations.

Automatically refreshing materialized views according to performance benefit

Match rules for rewriting queries to use materialized views may be dynamically generated by a database system. A database system may generate rules that indicate whether a given query can use a materialized view and how to rewrite the given query to use the materialized view. A query may be received and the rules may be applied to the query to determine that the query can use the materialized view and to rewrite the query to use the materialized view. The rewritten query can then be executed.

Dynamic generation of match rules for rewriting queries to use materialized views

A goal-oriented dialog system interacts with a user over one or more turns of dialog to determine a goal expressed by the user; the dialog system may then act to fulfill the goal by, for example, calling an application-programming interface. The user may supply dialog via text, speech, or other communication. The dialog system includes a first trained model, such as a translation model, to encode the dialog from the user into a context vector; a second trained model, such as another translation model, determines a plurality of candidate probabilities of items in a vocabulary. A language model determines responses to the user based on the input from the user, the context vector, and the plurality of candidate probabilities.

Goal-oriented dialog systems and methods

Techniques for intelligent task-oriented multi-turn dialog system automation are described. A seq2seq ML model can be trained using a corpus of training data and a loss function that is based at least in part on a distance to a goal. The seq2seq ML model can be provided a user utterance as an input, and a vector of a plurality of values output by a plurality of hidden units of a decoder of the seq2seq ML model can be used to select one or more candidate responses to the user utterance via a nearest neighbor algorithm. In some embodiments, the specially adapted seq2seq ML model can be trained using unsupervised learning, and can be adapted to select intelligent, coherent agent responses that move a task-oriented dialog toward its completion.

Task-oriented dialog systems utilizing combined supervised and reinforcement learning

Techniques described herein include a method and system for item demand forecasting that utilizes machine learning techniques to generate a set of quantiles. In some embodiments, several item features may be identified as being relevant to an item forecast and may be provided as inputs to a regression module, which may calculate a set of quantiles for each item. A set of quantiles may comprise a number of confidence levels or probabilities associated with calculated demand values for an item. In some embodiments, costs associated with the item may be used to select an appropriate quantile associated (e.g., based on a corresponding confidence level). In some embodiments, an item demand forecast may be generated based on the calculated demand value associated with the selected quantile. In some embodiments, one or more of the item may be automatically ordered based on that item demand forecast.

Demand forecasting via direct quantile loss optimization

Growing community of inventors

San Jose, CA, United States of America

Balakrishnan Narayanaswamy