### New talk in the LOS Seminar

Andra Băltoiu will give on Thursday, October 19, 2021 at 14:00 the talk
**Dictionary Learning Heuristics Control Through Adaptive Strategies and Applications to Anomaly Detection**
in the LOS seminar.

Andra Băltoiu will give on Thursday, October 19, 2021 at 14:00 the talk
**Dictionary Learning Heuristics Control Through Adaptive Strategies and Applications to Anomaly Detection**
in the LOS seminar.

Andrei Sipoş published the paper
**Quantitative inconsistent feasibility for averaged mappings**
in Optimization Letters.

Cristian Rusu and Lorenzo Rosasco published the paper
**Constructing fast approximate eigenspaces with application to the fast graph Fourier transforms**
in IEEE Transactions on Signal Processing.

The Research Center for Logic, Optimization and Security (LOS) was founded in October 2020 by
Laurenţiu Leuştean (head), Paul Irofti and Andrei Pătraşcu.

Our main objective is to stimulate interdisciplinary research in the fields of logic,
optimization and security. We are interested both in fundamental research as well as in
industrial applications. We focus on proof mining and applications to optimization, ergodic theory
and nonlinear analysis, convex optimization for machine learning, signal processing and
matrix factorization (dictionary learning), security, anomaly detection and anti-money laundry.

Contact

Hall 317, Faculty of Mathematics and Computer Science,

Academiei 14, 010014 Bucharest, Romania

Email: los@fmi.unibuc.ro

NetAlert aims to create a hardware-software sensor solution for detecting anomalies in computer networks
based on the monitoring and analysis of data packets.
The network-mounted sensor will provide real-time alerts on abnormal traffic behaviors
using two complementary approaches:

(i) static analysis based on rules and behavioral patterns;

(ii) machine learning (ML) analysis without prior expert knowledge.

The main goal of this project, called DDNET, is to adapt and propose
new dictionary learning methods for solving untractable
fault detection and isolation problems found in distribution networks.
Given a large dataset of sensor measurements from the distribution network,
the dictionary learning algorithms should be able to produce the subset
of network nodes where faults exist.

The proposed project, called Graphomaly, aims to create a
Python software package for anomaly detection in graphs
that model financial transactions,
with the purpose of discovering fraudulent behavior like money laundering,
illegal networks, tax evasion, scams, etc.
Such a toolbox is necessary in banks, where fraud detection departments
still use mostly human experts.

StOpAnomaly (143PD/2020)

StOpAnomaly aims to create, analyze and implement numerical optimization algorithms for large-scale optimization focusing on robust anomaly detection models based on decomposition and one-class classification. The research will be directed towards development of a toolbox containing scalable stochastic algorithms that can be used to detect several classes of anomalies in noisy large datasets.

Scientific seminars organized by LOS members

The working seminar of the LOS research center.

The logic seminar features talks on mathematical logic, philosophical logic and logical aspects of computer science.

The Cyber-security seminar brings together academic and industry folk to discuss hot topics in the field ranging from operating systems, static and dynamic analysis of executables to fraud detection and security centric machine learning techniques.

The Lean seminar is addressed to students who wish to master the Lean functional programming language.

- A. Sipoş, Quantitative inconsistent feasibility
for averaged mappings, Optimization Letters (2021),

DOI: 10.1007/s11590-021-01812-2. - C. Rusu, L. Rosasco,
Constructing fast approximate eigenspaces with application to the fast graph Fourier transforms,
IEEE Transactions on Signal Processing (2021),

DOI: 10.1109/TSP.2021.3107629. - A. Sipoş, Abstract strongly convergent variants of the proximal point algorithm , arXiv:2108.13994 [math.OC] (2021).
- A. Pătraşcu, P. Irofti, Computational complexity of Inexact Proximal Point Algorithm for Convex Optimization under Holderian Growth, arXiv:2108.04482 [cs.LG] (2021).
- A. Sipoş, A quantitative multi-parameter mean ergodic theorem , arXiv:2008.03932 [math.DS] (2021).
- L. Leuştean, H. Cheval, Quadratic rates of asymptotic regularity for the Tikhonov-Mann iteration, arXiv:2107.07176 [math.OC] (2021).
- A. Sipoş,
Construction of Fixed Points of Asymptotically Nonexpansive Mappings in Uniformly Convex Hyperbolic Spaces
, Numerical Functional Analysis and Optimization 42, 696-711 (2021),

DOI: 10.1080/01630563.2021.1924780. - A. Sipoş, Rates of metastability for iterations
on the unit interval, Journal of Mathematical Analysis and Applications 502, 125235 (2021),

DOI: 10.1016/j.jmaa.2021.125235. - A. Sipoş, Revisiting jointly firmly nonexpansive
families of mappings, Optimization (2021),

DOI: 10.1080/02331934.2021.1915312. - L. Leuştean, P. Pinto,
Quantitative results on a Halpern-type proximal point algorithm,
Computational Optimization and Applications 79, 101–125 (2021),

DOI: 10.1007/s10589-021-00263-w. - A. Pătraşcu, P. Irofti,
Stochastic proximal splitting algorithm for composite minimization,
Optimization Letters 15, 2255–2273 (2021),

DOI: 10.1007/s11590-021-01702-7. - C. Rusu, P. Irofti, Efficient and Parallel Separable Dictionary Learning, arXiv:2007.03800 [cs.LG] (2020).