US 8,572,071 B2Granted

Systems and methods for data transformation using higher order learning

AssigneeRutgers, The State University of New Jersey

InventorsWilliam M. Pottenger, Nikita Lytkin, Murat C. Ganiz

Life of this patent

Estimating the term…

Loading patent…

Abstract

Disclosed is a method and apparatus for transforming data in vector form. Each vector is composed of a set of attributes that are either boolean or have been mapped to boolean form. The vectors may or may not fall into categories assigned by a subject matter expert (SME). If categories exist, the categorical labels divide the vectors into subsets. The first transformation calculates a prior probability for each attribute based on the links between attributes in each subset of the vectors. The second transformation computes a new numeric value for each attribute based on the links between attributes in each subset of the vectors. The third transformation operates on vectors that have not been categorized. Based on the automatic selection of categories from the attributes, this transformation computes a new numeric value for each attribute based on the links between attributes in each subset of the vectors.

Claims22

Dependent claims

A method for transforming data comprising:

receiving, by a computing device, input data with categorical labels, the input data comprising vectors and each categorical label denoting a subset of the vectors;
forming, by the computing device, higher-order links for each subset of the vectors;
calculating, by the computing device, a higher-order transform for each attribute in each subset by calculating a higher-order prior for each attribute in each subset, wherein each higher-order prior represents a probability corresponding to the associated attribute that is based on one or more patterns of occurrence of the attribute, wherein calculating a higher-order prior comprises estimating the higher-order prior by:
P
^
⁡
(
x
i
=
1
|
X
)
=

φ
⁡
(
i
,
X
)


Φ
⁡
(
X
)

,
and
⁢
⁢
P
^
⁡
(
x
i
=
0
|
X
)
=
1
-
P
^
⁡
(
x
i
=
1
|
X
)
where Φ(X) denotes a set of higher-order links of a specified order in a dataset X, φ(i,X)⊂Φ(X) denotes a subset of higher-order links that contain attribute i in dataset X, set φ(i,X) defines an event that a randomly chosen higher-order link contains attribute i, sets Φ(X) and φ(i,X) allow for characterization of each attribute i by a probability mass function {circumflex over (P)}(xi|X) defined over two events:
presence of attribute i in a randomly chosen higher-order link, and an absence of that attribute from a randomly chosen higher-order link;
and
transmitting, by the computing device, transformed data to a learner to build a model.

61 more elements

A non-transitory computer readable storage medium tangibly storing computer program instructions capable of being executed by a computer processor, the computer program instructions defining the steps of:

receiving input data with categorical labels, the input data comprising vectors, and each categorical label denoting a subset of the vectors;
forming higher-order links for each subset of the vectors;
calculating, by the computing device, a higher-order transform for each attribute in each subset based on at least a portion of the higher-order priors by calculating a higher-order prior for each attribute in each subset, wherein each higher-order prior represents a probability corresponding to the associated attribute that is based on one or more patterns of occurrence of the attribute, wherein calculating a higher-order prior comprises estimating the higher-order prior by:
P
^
⁡
(
x
i
=
1
|
X
)
=

φ
⁡
(
i
,
X
)


Φ
⁡
(
X
)

,
and
⁢
⁢
P
^
⁡
(
x
i
=
0
|
X
)
=
1
-
P
^
⁡
(
x
i
=
1
|
X
)
where Φ(X) denotes a set of higher-order links of a specified order in a dataset X, φ(i,X)⊂Φ(X) denotes a subset of higher-order links that contain attribute i in dataset X, set φ(i,X) defines an event that a randomly chosen higher-order link contains attribute i, sets Φ(X) and φ(i,X) allow for characterization of each attribute i by a probability mass function {circumflex over (P)}(xi|X) defined over two events:
presence of attribute i in a randomly chosen higher-order link, and an absence of that attribute from a randomly chosen higher-order link;
and
transmitting transformed data to a learner to build a model.

61 more elements

Dependencies are inferred from the claim text.

Facts

Application: 12/797,366Filed 9 Jun 2010 · continuation in part
Priority: 19 Dec 2008
Published: US 2010/0312727 A19 Dec 2010
Granted: 29 Oct 2013
Term ends: Not estimatedThe source records don’t support an estimate.

Classification5 codes

G06N 20/00
Machine learning
G06F 18/2415
Relating to the classification model, e.g. parametric or non-parametric approaches: based on parametric or probabilistic models, e.g. based on…
G06F 18/254
Fusion techniques: of classification results, e.g. of results related to same input data
G06N
AI and machine learning

Family

Application12/317,134Parent of this continuation-in-part · filed 19 Dec 2008
Application61/185,255Us provisional application · filed 9 Jun 2009
US 2010/0312727 A1Pre-grant publication · published 9 Dec 2010
US 8,301,768 B2Parent of this continuation-in-part
US 2009/0177728 A1Published application
US 8,572,071 B2This patent

From the US filing history. Foreign family members aren’t listed.