Vijay S Iyengar

US Patent:

6351561, Feb 26, 2002

Filed:

Mar 26, 1999

Appl. No.:

09/276876

Inventors:

Vijay Sourirajan Iyengar - Cortlandt Manor NY

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06K 968

US Classification:

382226, 382228, 707102

Abstract:

A method and apparatus is disclosed for generating a decision tree classifier with oblique hyperplanes from a training set of records. The method iteratively comprises the steps of: initializing a set of vectors to the numeric attribute axes; creating a decision tree classifier using hyperplanes orthogonal to the set of vectors; checking if the iteration stopping criteria has been reached; computing a new set of vectors if the iteration proceeds; and choosing the best decision tree when the iteration is stopped. The vectors used are not restricted to the attribute axes and hence oblique hyperplanes are allowed to split nodes in the generated decision tree. The computation of the new vector set uses the decision tree produced in the latest iteration. The leaf nodes of this tree are considered pair-wise to compute the new vector set for use in the next iteration. The iterative process produces a set of decision trees from which the best one is chosen as the final result of the method.

US Patent:

6385607, May 7, 2002

Filed:

Mar 26, 1999

Appl. No.:

09/276870

Inventors:

Vijay Sourirajan Iyengar - Cortlandt Manor NY

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06F 1730

US Classification:

707 5, 707100

Abstract:

A method and apparatus is disclosed for generating a regression tree with oblique hyperplanes from a training set of records. The method is performed iteratively, stopping when a criterion has been reached. A new set of vectors is computed as the iteration proceeds. The vectors used are not restricted to the attribute axes and hence oblique hyperplanes are allowed to split nodes in the generated regression tree. Generally, the computation of the new vector set uses the regression tree produced in the latest iteration. The leaf nodes of this tree are considered pair-wise to compute the new vector set for use in the next iteration. The iterative process produces a set of regression trees from which a best tree is chosen as the final result of the method.

US Patent:

6721737, Apr 13, 2004

Filed:

Apr 4, 2001

Appl. No.:

09/826242

Inventors:

Vijay S. Iyengar - Cortlandt Manor NY
Jonathan Lee - Yorktown Heights NY

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06F 1730

US Classification:

707 5, 707 6, 707101, 709206, 705 7

Abstract:

A method is provided for ranking a plurality of items. The method comprises initializing a (D-1) dimensional weight space including a feasible region, where D is equal to a number of attributes and a point in the weight space corresponds to each attribute, determining an item pair, and querying a user to select an item from among the item pair. The method further includes reducing the feasible region based upon a users item selection, and ranking the items according a ranking point in a reduced feasible region. The ranking point is a center of the reduced feasible region, wherein the center is one of a vertex barycenter and center of gravity. The ranking point corresponds to a users item selection. The method includes the step of selecting a plurality of hyperplanes, each hyperplane corresponding to an item pair such that the hyperplane divides the feasible region into two substantially equal portions.

US Patent:

6937994, Aug 30, 2005

Filed:

May 16, 2000

Appl. No.:

09/571917

Inventors:

Vijay Sourirajan Iyengar - Cortlandt Manor NY, US

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06F017/60

US Classification:

705 10

Abstract:

A closed loop system is presented for selecting samples for labeling so that they can be used to generate classifiers. The sampling is done in phases. In each phase a subset of samples are chosen using information collected in previous phases and the classification model that has been generated up to that point. The total number of samples and the number of phases can be chosen by the user.

US Patent:

7024409, Apr 4, 2006

Filed:

Apr 16, 2002

Appl. No.:

10/123777

Inventors:

Vijay S. Iyengar - Cortlandt Manor NY, US

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06F 17/30

US Classification:

707 9, 707100, 707200

Abstract:

A data transform system comprises a processor, a memory connected to the processor, storing a collection of data, and a data transform module, accepting two data constraints and the collection of data from memory, wherein a first constraint is a usage constraint and a second constraint is a privacy constraint, the data transform module transforming the collection of data according to the usage constraint and the privacy constraint.

US Patent:

7295950, Nov 13, 2007

Filed:

Jun 23, 2005

Appl. No.:

11/159445

Inventors:

Vijay Sourirajan Iyengar - Cortlandt Manor NY, US

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06F 17/40

US Classification:

702182, 340500, 340679, 700 90, 700108, 702183, 702185, 702187

Abstract:

Provides methods, systems and apparatus for generating alerts for a system process that obtains raw channel data over time from one or more monitored channel of the system process. An example method includes processing the raw channel data to form time dependent signals based one or more user specified processing rules. The method produces alerts based on the deviation in behavior in one or more channels, where the deviation is quantified by a numeric level computed by comparing signals for varying time intervals with historically normal baseline signals. The method may include filtering the alerts to selectively form reportable alerts that are presented to the user based on user specified filtering rules.

US Patent:

7343272, Mar 11, 2008

Filed:

Feb 12, 2004

Appl. No.:

10/777548

Inventors:

Vijay S. Iyengar - Cortlandt Manor NY, US

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06G 7/48

US Classification:

703 6, 706 20

Abstract:

A system for detecting clusters in space and time using input data on occurrences of a phenomenon and characteristics at a plurality of locations and times comprises an expectation generation module determining expected occurrences of a phenomena, and an occurrence modeling module determining actual occurrences of the phenomena. The system further comprises a search module searching the expected occurrences and the actual occurrences for a plurality of candidate solutions, wherein each solution is represented as a set of points in the three-dimensional space, and wherein each point corresponds to a location at a time. The system comprises a convex container module determining at least one solution corresponding to a selected convex container shape from the plurality of candidate solutions, and a solution evaluation module determining a strength metric for each solution determined by the convex container module, the search module selecting a dominant cluster in the input data.

US Patent:

7792660, Sep 7, 2010

Filed:

Oct 26, 2007

Appl. No.:

11/925511

Inventors:

Vijay Sourirajan Iyengar - Cortlandt Manor NY, US

Assignee:

International Business Machines Corporation - Armonk NY

International Classification:

G06F 19/00
G06F 17/40

US Classification:

702183, 340679, 702182, 702187, 707758

Abstract:

Provides methods, systems and apparatus for generating alerts for a system process that obtains raw channel data over time from one or more monitored channel of the system process. An example method includes processing the raw channel data to form time dependent signals based one or more user specified processing rules. The method produces alerts based on the deviation in behavior in one or more channels, where the deviation is quantified by a numeric level computed by comparing signals for varying time intervals with historically normal baseline signals. The method may include filtering the alerts to selectively form reportable alerts that are presented to the user based on user specified filtering rules.