Methods and systems for dependency network analysis using a multitask learning graphical lasso objective function

Patent metadata

Field	Value
Publication number	US-9449284-B2
Application number	US-201314049891-A
Country	US
Kind code	B2
Filing date	Oct 9, 2013
Priority date	Oct 4, 2012
Publication date	Sep 20, 2016
Grant date	Sep 20, 2016

Abstract

Methods and systems for displaying dependencies within data and illustrating differences between a plurality of data sets are disclosed. In accordance with one such method, a plurality of data sets are received for the generation of a plurality of dependency networks in accordance with a graphical modeling scheme. The method further includes receiving a selection of a value of a parameter that adjusts a number of differences between the dependency networks in accordance with the graphical modeling scheme. In addition, at least one version of the dependency networks is generated based on the selected value of the parameter. Further, the one or more versions of the dependency networks is output to permit a user to analyze distinctions between the dependency networks.

First claim

What is claimed is:

1. A method for displaying dependencies within data and illustrating differences between a plurality of data sets comprising:
receiving the plurality of data sets for generation of a plurality of dependency networks in accordance with a graphical modeling scheme:
receiving a selection of a value of a parameter that adjusts a number of differences between the dependency networks in accordance with said graphical modeling scheme;
generating, by a hardware processor, at least one version of said dependency networks based on the selected value of the parameter;
outputting said at least one version of said dependency networks to permit a user to analyze distinctions between said dependency networks;
wherein said dependency networks are inferred together to facilitate comparison through the utilization of a multitask learning (MTL) transfer learning technique wherein the networks are biased to be similar to each other; and
wherein the graphical modeling scheme is based on a graphical lasso objective function that explicitly controls the number of differences learned and incorporates a transfer bias term.

2. The method of claim 1 , wherein said selection is a first selection, said first value is a first value, said at least one version is at least one first version and said method further comprises:
receiving a second selection of a second value of said parameter;
generating at least one second version of said dependency networks based on the selected second value of the parameter; and
outputting said at least one second version of said dependency networks to permit the user to analyze the distinctions between said dependency networks based on said at least one first version and of said at least one second version.

3. The method of claim 1 wherein said selection is a first selection, said value is a first value, said parameter is a first parameter and wherein said method further comprises:
receiving a second selection of a second value of a second parameter that adjusts a sparsity of edges within at least one of the dependency networks in accordance with said graphical modeling scheme, wherein said generating further comprises generating said at least one version of said dependency networks based on the selected second value of the second parameter.

4. The method of claim 3 , wherein said at least one version is at least one first version and said method further comprises:
receiving a third selection of a third value of said first parameter and a fourth selection of a fourth value of said second parameter;
generating at least one second version of said dependency networks based on the selected third value of the first parameter and the selected fourth value of said second parameter; and
outputting said at least one second version of said dependency networks to permit the user to analyze the distinctions between said dependency networks based on said at least one first version and said at least one second version.

5. The method of claim 1 , wherein said dependency networks are transelliptical graphical models.

6. The method of claim 5 , wherein said models are precision matrices and wherein each entry of each matrices denotes whether a dependency exists between two given variables.

7. The method according to claim 1 wherein for k classes of data, an estimate Σ k for
each set of data is made and a sparse precision matrix {umlaut over (Θ)} k is learned for each class of data by solving the following optimization function:

max

Θ
k

≻
0

,

∀
k

⁢

∑
k

⁢

⌊

log
⁢

⁢
det
⁢

⁢

Θ
k

-

t
⁢

⁢

r
⁡

(

∑
k

⁢

Θ
k

)

⌋

-

λ
1

⁡

(

1
-

λ
2

)

⁢


Θ


1

-

λ
1

⁢

λ
2

⁢

∑

i
≠
j

⁢



Θ

i
⁢

⁢
j



2

where ∥Θ∥ 1 is shorthand for entry wise L 1 norm of all Θ k , Θ ij is k-dimensional vector of partial correlations between data sets X i and X j , λ 1 controls the degree of sparsity, and λ 2 , 0≦λ 2 ≦1, that controls the number of differences learned.

8. A method for displaying dependencies within data and illustrating differences between a plurality of data sets comprising:
receiving the plurality of data sets for generation of a plurality of dependen

Assignees

• Nec Lab America Inc
• Nec Corp

Inventors

• Niculescu-Mizil Alexandru
• Oyen Diane

Classifications

• G16B45/00

  ICT specially adapted for bioinformatics-related data visualisation, e.g. displaying of maps or networks · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06N7/005

  Physics · mapped topic

• G06F19/26

  Physics · mapped topic