Sorry, you need to enable JavaScript to visit this website.

Learning theory and algorithms (MLR-LEAR)

Learning Flexible Representations of Stochastic Processes on Graphs


Graph convolutional networks adapt the architecture of convolutional neural networks to learn rich representations of data supported on arbitrary graphs by replacing the convolution operations of convolutional neural networks with graph-dependent linear operations. However, these graph-dependent linear operations are developed for scalar functions supported on undirected graphs. We propose both a generalization of the underlying graph and a class of linear operations for stochastic (time-varying) processes on directed (or undirected) graphs to be used in graph convolutional networks.

dsw.pdf

PDF icon poster (28 downloads)

Paper Details

Authors:
Brian M. Sadler, Radu V. Balan
Submitted On:
30 May 2018 - 1:35pm
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

poster

(28 downloads)

Subscribe

[1] Brian M. Sadler, Radu V. Balan, "Learning Flexible Representations of Stochastic Processes on Graphs", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3220. Accessed: Aug. 21, 2018.
@article{3220-18,
url = {http://sigport.org/3220},
author = {Brian M. Sadler; Radu V. Balan },
publisher = {IEEE SigPort},
title = {Learning Flexible Representations of Stochastic Processes on Graphs},
year = {2018} }
TY - EJOUR
T1 - Learning Flexible Representations of Stochastic Processes on Graphs
AU - Brian M. Sadler; Radu V. Balan
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3220
ER -
Brian M. Sadler, Radu V. Balan. (2018). Learning Flexible Representations of Stochastic Processes on Graphs. IEEE SigPort. http://sigport.org/3220
Brian M. Sadler, Radu V. Balan, 2018. Learning Flexible Representations of Stochastic Processes on Graphs. Available at: http://sigport.org/3220.
Brian M. Sadler, Radu V. Balan. (2018). "Learning Flexible Representations of Stochastic Processes on Graphs." Web.
1. Brian M. Sadler, Radu V. Balan. Learning Flexible Representations of Stochastic Processes on Graphs [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3220

Semi-Blind Inference of Topologies and Dynamical Processes over Graphs


Network science provides valuable insights across
numerous disciplines including sociology, biology, neuroscience
and engineering. A task of major practical importance in these
application domains is inferring the network structure from
noisy observations at a subset of nodes. Available methods for
topology inference typically assume that the process over the
network is observed at all nodes. However, application-specific
constraints may prevent acquiring network-wide observations.

Paper Details

Authors:
Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis
Submitted On:
29 May 2018 - 1:31pm
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

dsw_viysgg_18.pdf

(42 downloads)

Subscribe

[1] Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis, "Semi-Blind Inference of Topologies and Dynamical Processes over Graphs", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3214. Accessed: Aug. 21, 2018.
@article{3214-18,
url = {http://sigport.org/3214},
author = {Vassilis N. Ioannidis; Yanning Shen; Georgios B. Giannakis },
publisher = {IEEE SigPort},
title = {Semi-Blind Inference of Topologies and Dynamical Processes over Graphs},
year = {2018} }
TY - EJOUR
T1 - Semi-Blind Inference of Topologies and Dynamical Processes over Graphs
AU - Vassilis N. Ioannidis; Yanning Shen; Georgios B. Giannakis
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3214
ER -
Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis. (2018). Semi-Blind Inference of Topologies and Dynamical Processes over Graphs. IEEE SigPort. http://sigport.org/3214
Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis, 2018. Semi-Blind Inference of Topologies and Dynamical Processes over Graphs. Available at: http://sigport.org/3214.
Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis. (2018). "Semi-Blind Inference of Topologies and Dynamical Processes over Graphs." Web.
1. Vassilis N. Ioannidis, Yanning Shen, Georgios B. Giannakis. Semi-Blind Inference of Topologies and Dynamical Processes over Graphs [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3214

False Discovery Rate Control with Concave Penalties using Stability Selection


False discovery rate (FDR) control is highly desirable in several high-dimensional estimation problems. While solving such problems, it is observed that traditional approaches such as the Lasso select a high number of false positives, which increase with higher noise and correlation levels in the dataset. Stability selection is a procedure which uses randomization with the Lasso to reduce the number of false positives.

Paper Details

Authors:
Kush R. Varshney
Submitted On:
29 May 2018 - 1:22pm
Short Link:
Type:
Event:
Presenter's Name:
Document Year:
Cite

Document Files

Poster for FDR Control with Concave Penalties using Stability Selection

(30 downloads)

Subscribe

[1] Kush R. Varshney, "False Discovery Rate Control with Concave Penalties using Stability Selection", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3213. Accessed: Aug. 21, 2018.
@article{3213-18,
url = {http://sigport.org/3213},
author = {Kush R. Varshney },
publisher = {IEEE SigPort},
title = {False Discovery Rate Control with Concave Penalties using Stability Selection},
year = {2018} }
TY - EJOUR
T1 - False Discovery Rate Control with Concave Penalties using Stability Selection
AU - Kush R. Varshney
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3213
ER -
Kush R. Varshney. (2018). False Discovery Rate Control with Concave Penalties using Stability Selection. IEEE SigPort. http://sigport.org/3213
Kush R. Varshney, 2018. False Discovery Rate Control with Concave Penalties using Stability Selection. Available at: http://sigport.org/3213.
Kush R. Varshney. (2018). "False Discovery Rate Control with Concave Penalties using Stability Selection." Web.
1. Kush R. Varshney. False Discovery Rate Control with Concave Penalties using Stability Selection [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3213

On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization


Active graph-based semi-supervised learning (AG-SSL) aims to select a small set of labeled examples and utilize their graph-based relation to other unlabeled examples to aid in machine learning tasks. It is also closely related to the sampling theory in graph signal processing. In this paper, we revisit the original formulation of graph-based SSL and prove the supermodularity of an AG-SSL objective function under a broad class of regularization functions parameterized by Stieltjes matrices.

Paper Details

Authors:
Pin-Yu Chen, Dennis Wei
Submitted On:
20 April 2018 - 12:31am
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

poster

(66 downloads)

Subscribe

[1] Pin-Yu Chen, Dennis Wei, "On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3067. Accessed: Aug. 21, 2018.
@article{3067-18,
url = {http://sigport.org/3067},
author = {Pin-Yu Chen; Dennis Wei },
publisher = {IEEE SigPort},
title = {On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization},
year = {2018} }
TY - EJOUR
T1 - On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization
AU - Pin-Yu Chen; Dennis Wei
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3067
ER -
Pin-Yu Chen, Dennis Wei. (2018). On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization. IEEE SigPort. http://sigport.org/3067
Pin-Yu Chen, Dennis Wei, 2018. On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization. Available at: http://sigport.org/3067.
Pin-Yu Chen, Dennis Wei. (2018). "On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization." Web.
1. Pin-Yu Chen, Dennis Wei. On the Supermodularity of Active Graph-based Semi-supervised Learning with Stieltjes Matrix Regularization [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3067

THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE

Paper Details

Authors:
Submitted On:
19 April 2018 - 10:46pm
Short Link:
Type:
Event:
Presenter's Name:
Document Year:
Cite

Document Files

Slides-ChordDivergence18April2018.pdf

(51 downloads)

Subscribe

[1] , "THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3059. Accessed: Aug. 21, 2018.
@article{3059-18,
url = {http://sigport.org/3059},
author = { },
publisher = {IEEE SigPort},
title = {THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE},
year = {2018} }
TY - EJOUR
T1 - THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE
AU -
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3059
ER -
. (2018). THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE. IEEE SigPort. http://sigport.org/3059
, 2018. THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE. Available at: http://sigport.org/3059.
. (2018). "THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE." Web.
1. . THE CHORD GAP DIVERGENCE AND A GENERALIZATION OF THE BHATTACHARYYA DISTANCE [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3059

PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION

Paper Details

Authors:
Submitted On:
19 April 2018 - 2:13pm
Short Link:
Type:
Event:
Paper Code:
Document Year:
Cite

Document Files

presentation_1.pdf

(40 downloads)

Subscribe

[1] , "PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/2988. Accessed: Aug. 21, 2018.
@article{2988-18,
url = {http://sigport.org/2988},
author = { },
publisher = {IEEE SigPort},
title = {PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION},
year = {2018} }
TY - EJOUR
T1 - PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION
AU -
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/2988
ER -
. (2018). PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION. IEEE SigPort. http://sigport.org/2988
, 2018. PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION. Available at: http://sigport.org/2988.
. (2018). "PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION." Web.
1. . PARALLEL VECTOR FIELD REGULARIZED NON-NEGATIVE MATRIX FACTORIZATION FOR IMAGE REPRESENTATION [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/2988

A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations


In this paper we study the unmixing problem which aims
to separate a set of structured signals from their superposition.
In this paper, we consider the scenario in which
the mixture is observed via nonlinear compressive measurements.
We present a fast, robust, greedy algorithm called
Unmixing Matching Pursuit (UnmixMP) to solve this problem.
We prove rigorously that the algorithm can recover the
constituents from their noisy nonlinear compressive measurements
with arbitrarily small error. We compare our algorithm

Paper Details

Authors:
Dung Tran, Akshay Rangamani, Sang (Peter) Chin, Trac D. Tran
Submitted On:
17 April 2018 - 1:05pm
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

A Greedy Pursuit Algorithm For Separating Signals From.pdf

(64 downloads)

Subscribe

[1] Dung Tran, Akshay Rangamani, Sang (Peter) Chin, Trac D. Tran, "A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/2939. Accessed: Aug. 21, 2018.
@article{2939-18,
url = {http://sigport.org/2939},
author = {Dung Tran; Akshay Rangamani; Sang (Peter) Chin; Trac D. Tran },
publisher = {IEEE SigPort},
title = {A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations},
year = {2018} }
TY - EJOUR
T1 - A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations
AU - Dung Tran; Akshay Rangamani; Sang (Peter) Chin; Trac D. Tran
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/2939
ER -
Dung Tran, Akshay Rangamani, Sang (Peter) Chin, Trac D. Tran. (2018). A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations. IEEE SigPort. http://sigport.org/2939
Dung Tran, Akshay Rangamani, Sang (Peter) Chin, Trac D. Tran, 2018. A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations. Available at: http://sigport.org/2939.
Dung Tran, Akshay Rangamani, Sang (Peter) Chin, Trac D. Tran. (2018). "A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations." Web.
1. Dung Tran, Akshay Rangamani, Sang (Peter) Chin, Trac D. Tran. A Greedy Pursuit Algorithm For Separating Signals From Nonlinear Compressive Observations [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/2939

AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES


Given a data matrix with partially observed entries, the low-rank matrix completion problem is one of finding a matrix with the lowest rank that perfectly fits the given observations. While there exist convex relaxations for the low-rank completion problem, the underlying problem is inherently non-convex, and most algorithms (alternating projection, Riemannian optimization, etc.) heavily depend on the initialization. This paper proposes an improved initialization that relies on successive rank-1 updates.

Paper Details

Authors:
Ahmed Douik, Babak Hassibi
Submitted On:
19 April 2018 - 3:10pm
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

Presentation_v2.pdf

(56 downloads)

Presentation Slides

(61 downloads)

Subscribe

[1] Ahmed Douik, Babak Hassibi, "AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/2921. Accessed: Aug. 21, 2018.
@article{2921-18,
url = {http://sigport.org/2921},
author = {Ahmed Douik; Babak Hassibi },
publisher = {IEEE SigPort},
title = {AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES},
year = {2018} }
TY - EJOUR
T1 - AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES
AU - Ahmed Douik; Babak Hassibi
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/2921
ER -
Ahmed Douik, Babak Hassibi. (2018). AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES. IEEE SigPort. http://sigport.org/2921
Ahmed Douik, Babak Hassibi, 2018. AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES. Available at: http://sigport.org/2921.
Ahmed Douik, Babak Hassibi. (2018). "AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES." Web.
1. Ahmed Douik, Babak Hassibi. AN IMPROVED INITIALIZATION FOR LOW-RANK MATRIX COMPLETION BASED ON RANK-1 UPDATES [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/2921

Improved Algorithms for Differentially private Orthogonal Tensor Decomposition


Tensor decompositions have applications in many areas including signal processing, machine learning, computer vision and neuroscience. In this paper, we propose two new differentially private algorithms for orthogonal decomposition of symmetric tensors from private or sensitive data; these arise in applications such as latent variable models. Differential privacy is a formal privacy framework that guarantees protections against adversarial inference.

Paper Details

Authors:
Submitted On:
13 April 2018 - 2:23pm
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

Presentation slides

(50 downloads)

Subscribe

[1] , "Improved Algorithms for Differentially private Orthogonal Tensor Decomposition", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/2743. Accessed: Aug. 21, 2018.
@article{2743-18,
url = {http://sigport.org/2743},
author = { },
publisher = {IEEE SigPort},
title = {Improved Algorithms for Differentially private Orthogonal Tensor Decomposition},
year = {2018} }
TY - EJOUR
T1 - Improved Algorithms for Differentially private Orthogonal Tensor Decomposition
AU -
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/2743
ER -
. (2018). Improved Algorithms for Differentially private Orthogonal Tensor Decomposition. IEEE SigPort. http://sigport.org/2743
, 2018. Improved Algorithms for Differentially private Orthogonal Tensor Decomposition. Available at: http://sigport.org/2743.
. (2018). "Improved Algorithms for Differentially private Orthogonal Tensor Decomposition." Web.
1. . Improved Algorithms for Differentially private Orthogonal Tensor Decomposition [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/2743

Differentially private Distributed Principal Component Analysis


Differential privacy is a cryptographically-motivated formal privacy definition that is robust against strong adversaries. The principal component analysis (PCA) algorithm is frequently used in signal processing, machine learning, and statistics pipelines. In many scenarios, private or sensitive data is distributed across different sites: in this paper we propose a differentially private distributed PCA scheme to enable collaborative dimensionality reduction.

Paper Details

Authors:
Submitted On:
13 April 2018 - 2:20pm
Short Link:
Type:
Event:
Presenter's Name:
Paper Code:
Document Year:
Cite

Document Files

Presentation slides

(83 downloads)

Subscribe

[1] , "Differentially private Distributed Principal Component Analysis", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/2742. Accessed: Aug. 21, 2018.
@article{2742-18,
url = {http://sigport.org/2742},
author = { },
publisher = {IEEE SigPort},
title = {Differentially private Distributed Principal Component Analysis},
year = {2018} }
TY - EJOUR
T1 - Differentially private Distributed Principal Component Analysis
AU -
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/2742
ER -
. (2018). Differentially private Distributed Principal Component Analysis. IEEE SigPort. http://sigport.org/2742
, 2018. Differentially private Distributed Principal Component Analysis. Available at: http://sigport.org/2742.
. (2018). "Differentially private Distributed Principal Component Analysis." Web.
1. . Differentially private Distributed Principal Component Analysis [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/2742

Pages