TY  - CHAP
AU  - Szeghy, David
AU  - Aslan, Mahmoud
AU  - Fóthi, Áron
AU  - Meszaros, Balazs
AU  - Milacski, Zoltán Ádám
AU  - Lőrincz, András
ED  - Fred, A
ED  - Sansone, C
ED  - Gusikhin, O
ED  - Madani, K
TI  - Structural Extensions of Basis Pursuit: Guarantees on Adversarial Robustness
T2  - DELTA: PROCEEDINGS OF THE 3RD INTERNATIONAL CONFERENCE ON DEEP LEARNING THEORY AND APPLICATIONS
PB  - SciTePress
CY  - Setubal
SN  - 9789897585845
PY  - 2022
SP  - 77
EP  - 85
PG  - 9
DO  - 10.5220/0011138900003277
UR  - https://m2.mtmt.hu/api/publication/33305314
ID  - 33305314
AB  - While deep neural networks are sensitive to adversarial noise, sparse coding using the Basis Pursuit (BP) method is robust against such attacks, including its multi-layer extensions. We prove that the stability theorem of BP holds upon the following generalizations: (i) the regularization procedure can be separated into disjoint groups with different weights, (ii) neurons or full layers may form groups, and (iii) the regularizer takes various generalized forms of the l(1) norm. This result provides the proof for the architectural generalizations of (Cazenavette et al., 2021) including (iv) an approximation of the complete architecture as a shallow sparse coding network. Due to this approximation, we settled to experimenting with shallow networks and studied their robustness against the Iterative Fast Gradient Sign Method on a synthetic dataset and MNIST. We introduce classification based on the l(2) norms of the groups and show numerically that it can be accurate and offers considerable speedups. In this family, linear transformer shows the best performance. Based on the theoretical results and the numerical simulations, we highlight numerical matters that may improve performance further. The proofs of our theorems can be found in the supplementary material*.
LA  - English
DB  - MTMT
ER  - 

TY  - GEN
AU  - Szeghy, Dávid
AU  - Mahmoud, Aslan
AU  - Fóthi, Áron
AU  - Balázs, Mészáros
AU  - Milacski, Zoltán Ádám
AU  - Lőrincz, András
TI  - Structural Extensions of Basis Pursuit: Guarantees on Adversarial Robustness
PY  - 2022
SP  - 77
EP  - 85
PG  - 20
UR  - https://m2.mtmt.hu/api/publication/32847308
ID  - 32847308
LA  - English
DB  - MTMT
ER  - 

TY  - THES
AU  - Milacski, Zoltán Ádám
TI  - "Temporal Reconstruction Methods in Self-Supervised Machine Learning
PB  - Eötvös Loránd Tudományegyetem (ELTE)
PY  - 2021
SP  - 170
DO  - 10.15476/ELTE.2021.018
UR  - https://m2.mtmt.hu/api/publication/32643465
ID  - 32643465
LA  - English
DB  - MTMT
ER  - 

TY  - JOUR
AU  - Fodor, Ádám
AU  - Kopácsi, László
AU  - Milacski, Zoltán Ádám
AU  - Lőrincz, András
TI  - Speech de-identification with deep neural networks
JF  - ACTA CYBERNETICA
J2  - ACTA CYBERN-SZEGED
VL  - 25
PY  - 2021
IS  - 2
SP  - 257
EP  - 269
PG  - 13
SN  - 0324-721X
DO  - 10.14232/actacyb.288282
UR  - https://m2.mtmt.hu/api/publication/32538324
ID  - 32538324
LA  - English
DB  - MTMT
ER  - 

TY  - JOUR
AU  - Han, Changhee
AU  - Rundo, Leonardo
AU  - Murao, Kohei
AU  - Noguchi, Tomoyuki
AU  - Shimahara, Yuki
AU  - Milacski, Zoltán Ádám
AU  - Koshino, Saori
AU  - Sala, Evis
AU  - Nakayama, Hideki
AU  - Satoh, Shin’ichi
TI  - MADGAN. unsupervised medical anomaly detection GAN using multiple adjacent brain MRI slice reconstruction
TS  - unsupervised medical anomaly detection GAN using multiple adjacent brain MRI slice reconstruction
JF  - BMC BIOINFORMATICS
J2  - BMC BIOINFORMATICS
VL  - 22
PY  - 2021
IS  - S2
SN  - 1471-2105
DO  - 10.1186/s12859-020-03936-1
UR  - https://m2.mtmt.hu/api/publication/32252538
ID  - 32252538
LA  - English
DB  - MTMT
ER  - 

TY  - GEN
AU  - Dávid, Szeghy
AU  - Milacski, Zoltán Ádám
AU  - Fóthi, Áron
AU  - Lőrincz, András
TI  - Adversarial Perturbation Stability of the Layered Group Basis Pursuit
PY  - 2021
UR  - https://m2.mtmt.hu/api/publication/32153563
ID  - 32153563
N1  - Conference on Mathematics of Machine Learning, August 04 -07, 2021, Center for Interdisciplinary Research (ZiF), Bielefeld
LA  - English
DB  - MTMT
ER  - 

TY  - JOUR
AU  - Fóthi, Áron
AU  - Faragó, Kinga Bettina
AU  - Kopácsi, László
AU  - Milacski, Zoltán Ádám
AU  - Varga, Viktor
AU  - Lőrincz, András
TI  - Multi Object Tracking for Similar Instances. A Hybrid Architecture
TS  - A Hybrid Architecture
JF  - LECTURE NOTES IN COMPUTER SCIENCE
J2  - LNCS
VL  - 12532
PY  - 2020
SP  - 436
EP  - 447
PG  - 12
SN  - 0302-9743
DO  - 10.1007/978-3-030-63830-6_37
UR  - https://m2.mtmt.hu/api/publication/31709217
ID  - 31709217
LA  - English
DB  - MTMT
ER  - 

TY  - GEN
AU  - Changhee, Han
AU  - Leonardo, Rundo
AU  - Kohei, Murao
AU  - Tomoyuki, Noguchi
AU  - Yuki, Shimahara
AU  - Milacski, Zoltán Ádám
AU  - Saori, Koshino
AU  - Evis, Sala
AU  - Hideki, Nakayama
AU  - Shinichi, Satoh
TI  - MADGAN: unsupervised Medical Anomaly Detection GAN using multiple adjacent brain MRI slice reconstruction
PY  - 2020
UR  - https://m2.mtmt.hu/api/publication/31565402
ID  - 31565402
N1  - https://arxiv.org/pdf/2007.13559.pdf
LA  - English
DB  - MTMT
ER  - 

TY  - JOUR
AU  - Milacski, Zoltán Ádám
AU  - Póczos, Barnabás
AU  - Lőrincz, András
TI  - VideoOneNet. Bidirectional Convolutional Recurrent OneNet with Trainable Data Steps for Video Processing
TS  - Bidirectional Convolutional Recurrent OneNet with Trainable Data Steps for Video Processing
JF  - PROCEEDINGS OF MACHINE LEARNING RESEARCH
J2  - PROC MACH LEAR RESEARCH
VL  - 119
PY  - 2020
SP  - 6893
EP  - 6904
PG  - 12
SN  - 2640-3498
UR  - https://m2.mtmt.hu/api/publication/31565326
ID  - 31565326
LA  - English
DB  - MTMT
ER  - 

TY  - CONF
AU  - Fodor, Ádám
AU  - Kopácsi, László
AU  - Milacski, Zoltán Ádám
AU  - Lőrincz, András
TI  - Speech de-identification with deep neural networks
T2  - The 12th Conference of PhD Students in Computer Science
PB  - Szegedi Tudományegyetem (SZTE)
C1  - Szeged
PY  - 2020
SP  - 7
EP  - 10
PG  - 4
UR  - https://m2.mtmt.hu/api/publication/31523907
ID  - 31523907
LA  - English
DB  - MTMT
ER  -