Oracle International Corporation

Subhransu Basu

Angelo Pruscino

Michael Zoll

Shrikanth Shankar

Wataru Miyoshi

Yaser I Suleiman

Thomas Herter

Thomas Breidt

Klaus Thielen

Sahil Kumar

Wolfgang Lohwasser

Ananth Raghavan

Vikram Shukla

Harmeek Singh Bedi

Russell John Green

Kumar Rajamani

Jaebock Lee

Ho Chak Hung

Described is an approach that provides an adaptive solution to missing data for machine learning systems. A gradient solution is provided that is attentive to imputation needs at each of several missingness levels. This multilevel approach treats data missingness at any of multiple severity levels while utilizing, as much as possible, the actual observed data.

Method and system for adaptively imputing sparse and missing data for predictive models

Described is an improved approach to remove data outliers by filtering out data correlated to detrimental events within a system. One or more detrimental even conditions are defined to identify and handle abnormal transient states from collected data for a monitored system.

Method and system for adaptively removing outliers from data used in training of predictive models

Described is an approach for performing context-aware prognoses in machine learning systems. The approach harnesses streams of detailed data collected from a monitored target to create a context, in parallel to ongoing model operations, for the model outcomes. The context is then probed to identify the particular elements associated with the model findings.

Method and system for performing context-aware prognoses for health analysis of monitored systems

Described is an improved approach to implement selection of training data for machine learning, by presenting a designated set of specific data indicators where these data indicators correspond to metrics that end users are familiar with and are easily understood by ordinary users and DBAs within their knowledge domain. Selection of these indicators would correlate automatically to selection of a corresponding set of other metrics/signals that are less understandable to an ordinary user. Additional analysis of the selected data can then be performed to identify and correct any statistical problems with the selected training data.

Method and system for cleansing training data for predictive models

Techniques are described herein for partitioning a database table using a virtual column as the partitioning key. A virtual column is defined by the database to contain results of a defining expression. Partitioning can be performed using a variety of partitioning schemes, such as range partitioning and hash partitioning. The partitions are automatically updated to reflect changes to the values in the base columns upon which the virtual column depends. Partition pruning and parallel-wise partition joins can be performed on database tables which have been partitioned on virtual columns.

Partitioning in virtual columns

Techniques are described herein for performing column functions on virtual columns in database tables. A virtual column is defined by the database to contain results of a defining expression. Statistics are collected and maintained for virtual columns. Indexing is performed on virtual columns. Referential integrity is maintained between two tables using virtual columns as keys. Join predicate push-down operations are also performed using virtual columns.

Virtual columns

Techniques are described herein for replacing a part of an expression in a query to a database with a virtual column. A virtual column is defined by the database to contain results of a defining expression. A part of an expression in a query is replaced with a virtual column whose defining expression is equal to the part of the expression in the query. The part of an expression in a query can be the entire expression. When an expression in a query can be replaced with more than one replacement expression involving virtual columns, the replacement expression with the least cost of query processing is selected.

Expression replacement in virtual columns

Versioned metadata allows multiple versions of metadata for a given object to exist concurrently on a system thereby supporting the parallel execution of operations which would otherwise be mutually exclusive. Metadata updates are permitted while other objects maintain access to another version of the metadata through a versioning infrastructure. The versioning infrastructure allows the creation of new versions of metadata, maintains obsolete versions in the system as long as they are being used by any object, and deletes the obsolete versions once they are no longer in use.

System and method for managing versions of metadata

The concept of portioning is expanded with a variety of techniques. In particular, one technique involves partitioning an object at multiple levels, where at least one of the levels uses list-based partitioning. Further, a partitioning technique is provided which involves storing a data item in a default partition when the item does not map to any one of the list-based partitions for which a key value group has been defined. Techniques are also provided for efficiently splitting a partition.

Methods for partitioning an object

A system, method, computer program and article of manufacture for sharing metadata among cursors is provided. A metadata object marked as a versioned object can have multiple versions. Program units can share these versions of object metadata as opposed to copying the non-versioned object metadata into their memory. These program units may contain a pointer to an object's metadata but are otherwise void of the metadata information for the object. A locking protocol for the metadata ensures concurrency control and triggers dependency tracking. A locking protocol for the cursors controls cursor dependency status.

System for allowing object metadata to be shared between cursors for concurrent read write access

The concept of partitioning is expanded with a variety of techniques. In particular, one technique involves partitioning an object at multiple levels, where at least one of the levels uses list-based partitioning. Further, a partitioning technique is provided which involves storing a data item in a default partition when the item does not map to any one of the list-based partitions for which a key value group has been defined. Techniques are also provided for efficiently splitting a partition.

Growing community of inventors

Fremont, CA, United States of America

Subhransu Basu