Change search
Link to record
Permanent link

Direct link
BETA
Publications (5 of 5) Show all publications
Liu, D., Vu, M. T., Chatterjee, S. & Rasmussen, L. K. (2019). ENTROPY-REGULARIZED OPTIMAL TRANSPORT GENERATIVE MODELS. In: 2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP): . Paper presented at 44th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), MAY 12-17, 2019, Brighton, ENGLAND (pp. 3532-3536). IEEE
Open this publication in new window or tab >>ENTROPY-REGULARIZED OPTIMAL TRANSPORT GENERATIVE MODELS
2019 (English)In: 2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), IEEE , 2019, p. 3532-3536Conference paper, Published paper (Refereed)
Abstract [en]

We investigate the use of entropy-regularized optimal transport (EOT) cost in developing generative models to learn implicit distributions. Two generative models are proposed. One uses EOT cost directly in an one-shot optimization problem and the other uses EOT cost iteratively in an adversarial game. The proposed generative models show improved performance over contemporary models on scores of sample based test.

Place, publisher, year, edition, pages
IEEE, 2019
Series
International Conference on Acoustics Speech and Signal Processing ICASSP, ISSN 1520-6149
Keywords
Optimal transport, generative models
National Category
Computer and Information Sciences
Identifiers
urn:nbn:se:kth:diva-261047 (URN)10.1109/ICASSP.2019.8682721 (DOI)000482554003151 ()2-s2.0-85068999197 (Scopus ID)978-1-4799-8131-1 (ISBN)
Conference
44th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), MAY 12-17, 2019, Brighton, ENGLAND
Note

QC 20191001

Available from: 2019-10-01 Created: 2019-10-01 Last updated: 2019-10-01Bibliographically approved
Vu, M. T. (2019). Perspectives on Identification Systems. (Doctoral dissertation). Stockholm: KTH Royal Institute of Technology
Open this publication in new window or tab >>Perspectives on Identification Systems
2019 (English)Doctoral thesis, monograph (Other academic)
Abstract [en]

Identification systems such as biometric identification systems have been becoming ubiquitous. Fundamental bounds on the performance of the systems have been established in literature. In this thesis we further relax several assumptions in the identification problem and derive the corresponding fundamental regions for these settings.

The generic identification architecture is first extended so that users’ information is stored in two layers. Additionally, the processing is separated in two steps where the observation sequence in the first step is a noisy, pre-processed version of the original one. This setting generalizes several known settings in the literature. Given fixed pre-processing schemes, we study optimal trade-offs in the discrete and Gaussian cases. As corollaries we also provide characterizations for related problems.

In a second aspect, the joint distribution in the identification problem is relaxed in several ways. We first assume that all users’ sequences are drawn from a common distribution, which depends on a state of the system. The observation sequence is induced by a channel which has its own state. Another variant, in which the channel is fixed, however the distributions of users’ sequences are not necessarily identical, is considered next. We then study the case that users’ data sequence are generated independently from a mixture distribution. Optimal performance regions of these settings are provided. We further give an inner bound and an outer bound on the region when the observation channel varies arbitrarily. Additionally, we strengthen the relation between the Wyner-Ahlswede-Körner problem and the identification problem and show the equivalence of these two.

Finally, we study a binary hypothesis testing problem which decides whether or not the observation sequence is related to one user in the database. The optimal exponent of the second type of error is studied. Furthermore, we show that the single-user testing against independence problem studied by Ahlswede and Csiszár is equivalent to the identification problem as well as the Wyner-Ahlswede-Körner problem.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2019. p. 151
Series
TRITA-EECS-AVL ; 2019:57
National Category
Electrical Engineering, Electronic Engineering, Information Engineering
Research subject
Electrical Engineering
Identifiers
urn:nbn:se:kth:diva-254611 (URN)978-91-7873-239-5 (ISBN)
Public defence
2019-08-29, F3, Lindstedtsvägen 26, Stockholm, 13:15 (English)
Opponent
Supervisors
Note

QC 20190708

Available from: 2019-07-08 Created: 2019-07-02 Last updated: 2019-07-08Bibliographically approved
Vu, M. T., Oechtering, T. J. & Skoglund, M. (2019). Testing in identification systems. In: 2018 IEEE Information Theory Workshop, ITW 2018: . Paper presented at 2018 IEEE Information Theory Workshop, ITW 2018, 25 November 2018 through 29 November 2018. Institute of Electrical and Electronics Engineers Inc.
Open this publication in new window or tab >>Testing in identification systems
2019 (English)In: 2018 IEEE Information Theory Workshop, ITW 2018, Institute of Electrical and Electronics Engineers Inc. , 2019Conference paper, Published paper (Refereed)
Abstract [en]

We study a hypothesis testing problem to decide whether or not an observation sequence is related to one of users in a database which contains compressed versions of users' data. Our main interest lies on the characterization of the exponent of the probability of the second kind of error when the number of users in the database grows exponentially. We show a lower bound on the error exponent and identify special cases where the bound is tight. Next, we study the -achievable error exponent and show a sub-region where the lower bound is tight.

Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers Inc., 2019
Keywords
Information-spectrum method, Mixture distribution, Strong converse, Errors, Error exponent, Hypothesis testing, Information spectrum, Lower bounds, Mixture distributions, Second kinds, Sub-regions, Information theory
National Category
Electrical Engineering, Electronic Engineering, Information Engineering
Identifiers
urn:nbn:se:kth:diva-248277 (URN)10.1109/ITW.2018.8613310 (DOI)2-s2.0-85062066143 (Scopus ID)9781538635995 (ISBN)
Conference
2018 IEEE Information Theory Workshop, ITW 2018, 25 November 2018 through 29 November 2018
Note

QC 20190405

Available from: 2019-04-05 Created: 2019-04-05 Last updated: 2019-04-05Bibliographically approved
Vu, M. T., Oechtering, T. J. & Skoglund, M. (2018). Gaussian hierarchical identification with pre-processing. In: Data Compression Conference Proceedings: . Paper presented at 2018 Data Compression Conference, DCC 2018, 27 March 2018 through 30 March 2018 (pp. 277-286). Institute of Electrical and Electronics Engineers Inc.
Open this publication in new window or tab >>Gaussian hierarchical identification with pre-processing
2018 (English)In: Data Compression Conference Proceedings, Institute of Electrical and Electronics Engineers Inc. , 2018, p. 277-286Conference paper, Published paper (Refereed)
Abstract [en]

In this work we consider a two-stage identification problem with pre-processing where the users' data and observation are Gaussian distributed. In the first stage the processing unit returns a list of compatible users using the information from the first storage layer and the pre-processed observation. Then, the refined search is performed in the second stage where the processing unit returns the exact user's identity and a corresponding reconstruction sequence. We provide a complete rate-distortion trade-off for the Gaussian setting.

Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers Inc., 2018
Keywords
Compression, Identification system, Compaction, Digital storage, Economic and social effects, Electric distortion, Gaussian distribution, Image coding, Signal distortion, Gaussian distributed, Gaussian setting, Hierarchical identification, Pre-processing, Processing units, Rate distortion trade-off, Storage layers, Two-stage identification, Data compression
National Category
Computer and Information Sciences
Identifiers
urn:nbn:se:kth:diva-238076 (URN)10.1109/DCC.2018.00036 (DOI)2-s2.0-85050965248 (Scopus ID)9781538648834 (ISBN)
Conference
2018 Data Compression Conference, DCC 2018, 27 March 2018 through 30 March 2018
Note

Conference code: 138136; Export Date: 30 October 2018; Conference Paper; CODEN: DDCCF

QC 20190114

Available from: 2019-01-14 Created: 2019-01-14 Last updated: 2019-01-14Bibliographically approved
Vu, M. T., Oechtering, T. J. & Skoglund, M. (2018). Testing in Identification Systems. In: 2018 IEEE INFORMATION THEORY WORKSHOP (ITW): . Paper presented at IEEE Information Theory Workshop (ITW), NOV 25-29, 2018, Guangzhou, PEOPLES R CHINA (pp. 295-299). IEEE
Open this publication in new window or tab >>Testing in Identification Systems
2018 (English)In: 2018 IEEE INFORMATION THEORY WORKSHOP (ITW), IEEE , 2018, p. 295-299Conference paper, Published paper (Refereed)
Abstract [en]

We study a hypothesis testing problem to decide whether or not an observation sequence is related to one of users in a database which contains compressed versions of users' data. Our main interest lies on the characterization of the exponent of the probability of the second kind of error when the number of users in the database grows exponentially. We show a lower bound on the error exponent and identify special cases where the bound is tight. Next, we study the c-achievable error exponent and show a sub-region where the lower bound is tight.

Place, publisher, year, edition, pages
IEEE, 2018
Series
Information Theory Workshop, ISSN 2475-420X
Keywords
Mixture distribution, information-spectrum method, strong converse
National Category
Computer and Information Sciences
Identifiers
urn:nbn:se:kth:diva-252670 (URN)10.1109/ITW.2018.8613310 (DOI)000467849900060 ()2-s2.0-85062066143 (Scopus ID)978-1-5386-3599-5 (ISBN)
Conference
IEEE Information Theory Workshop (ITW), NOV 25-29, 2018, Guangzhou, PEOPLES R CHINA
Note

QC 20190610

Available from: 2019-06-10 Created: 2019-06-10 Last updated: 2019-06-10Bibliographically approved
Organisations
Identifiers
ORCID iD: ORCID iD iconorcid.org/0000-0002-0737-2531

Search in DiVA

Show all publications