Microsoft Technology Licensing, LLC

Vijaykumar K Aski

Danny Chen

Neil Tenenholtz

Neil Evan Lydick

Nikolay Sergeyevich Rovinskiy

Dhawal Dilip Parkar

Nazmiye Ceren Abay

The disclosed techniques pertain to the dynamic control of select functions that are applied to a time series dataset based on the detection of stationary time series grains. In some configurations, a system selectively applies select functions, e.g., the application of a differencing function, to a dataset in response to determining that a number of stationary time series grains detected in the dataset meets one or more criteria with respect to a threshold. If a system determines that the number of stationary time series grains meets one or more criteria with respect to a threshold, the system can apply a differencing function to the entire dataset. By controlling the differencing function based on the detection of stationary time series grains with respect to a threshold, the system can increase the accuracy and the efficiency of a machine learning system or any other system that utilizes time series datasets.

Dynamically controlled processing of time series datasets based on the detection of stationary time series grains

Techniques for implementing a low-latency query processor accommodating an arbitrary number of data rows with no column indexing. In an aspect, data is stored across a plurality of component databases, with no requirement to strictly allocate data to partitions based on row keys. A histogram table is provided to map object relationships identified in a user query to the component databases where relevant data is stored. A server processing the user query communicates with component databases via an intermediary module. The intermediary module may include intermediary nodes dynamically assigned to connect to the component databases to retrieve and process the queried data.

Low-latency query processor

A mapping is received and stored that maps elements of a data warehouse to types of a type system implemented by a data source. Program code is generated that performs a transform of data retrieved from a data source based on the mapping. Generation of the program code may include generating program code for performing a dimension transform based on the mapping, generating program code for performing a fact transform based on the mapping, and generating program code for performing an outrigger transform based on the mapping. The generated program code may then be executed to transform the data retrieved from the data source prior to loading into the data warehouse.

Growing community of inventors

Bellevue, WA, United States of America

Vijaykumar K Aski