Science seminar

Seminars are held online. To subscribe to the newsletter, write to the lab manager kzherebczova@hse.ru.

If you would like to be a speaker at lab seminar, please send your topic and brief description to kzherebczova@hse.ru

22.11.2022 Анализ академической успеваемости студентов с использованием журналов событий электронной образовательной среды

Докладчик: Шаимов Никита, стажер-исследователь

Доклад посвящен исследованию данных из электронной образовательной среды при помощи методов интеллектуального анализа процессов (Process Mining). В связи с процессом цифровизации в образовании все чаще внедряются различные системы управления обучением, а документы и ведомости переходят в электронный формат. Данное исследование предлагает подход анализу данных из электронных систем с точки зрения образовательных процессов для поиска проблемных мест и ошибок при построении образовательных программ. Анализ позволяет администраторам образовательного процесса эффективно использовать имеющиеся массивы данных для повышения качества образования. В рамках доклада будут рассмотрены теоретические основы подхода, особенности подготовки используемых данных, и применение методов на реальных образовательных данных.

15.09.2022 Overview of real-time models

Speaker: Evgenii Vinarskii, PAIS Lab Research Assistant 

This presentation is devoted to the current deals in studying real-time models. We consider the classical Timed Automaton (TA) model. Then we move to the Timed Finite State Machines (TFSM) model. There are three main TFSM models: (i) TFSM with timed guards, (ii) TFSM with timeouts (iii) TFSM with timed guards and timeouts. We consider equivalence relation between various real-time models and bisimulation relation between timed and untimed models.

08.09.2022 Studying the properties of timed finite state machines

Speaker: Evgenii Vinarskii, PAIS Lab Research Assistant 

Sequential reactive systems are formal models of programs that interact with the environment by receiving inputs and producing corresponding outputs. The behavior of a reactive system depends not only on the flow of input data, but also on the time the input data arrive and the delays that occur when generating responses. To capture these aspects, a timed finite state machine (TFSM) is used. TFSM is an extension of finite state machine (FSM) which is used for modelling and analysis of real-time reactive systems. 

This presentation is devoted to refined semantics ofTFSMs, this gives us an opportunity to model more complex real-time systems.For example, we show how to represent the behavior of Software Defined Networking (SDN) Controller via TFSM. Then we discuss how the verification methods for the refined TFSM change compared with classical one.

21.01.2022 Data Petri Nets as a Formalism for Representing Decision-Aware Process Models

Speaker: Nikolay Suvorov, PAIS Lab Research Assistant

In recent years, there has been an increasing interest in enriching the traditional control flow perspective of processes with additional dimensions. This report is focused on Data Petri nets (DPNs) which are a data aware extension of Petri nets where the net is enriched with data variables of different types, while transitions are guarded by formulae that inspect and update such variables. Such formalism provides a possibility to consider not only the control flow perspective of the process but also the dataflow perspective as well as the decisions made based on the data.

The report covers syntax and semantics of DPNs, techniques to discover DPNs from the event data, ways to perform conformance checking on models represented as DPNs and different algorithms to check soundness of DPNs.

17.06.2021 Object-Centric Replay-Based Conformance Checking: Unveiling Desire Lines and Local Deviations

Speaker: Julio Cesar Carrasquel Gamez, PAIS Lab Research Assistant 

Conformance checking methods diagnose whether a process, as observed in an event log, complies with its specification model, e.g., a Petri net. Most of these methods, however, check process instances (cases) individually, neglecting the interaction of such instances in a system. This limitation has been addressed in other areas of process mining by object-centric approaches. These approaches make use of Petri net extensions to model and analyze overlapping instances that are centered on the handling of objects. For instance, some of these approaches were recently presented at the PAIS Lab's seminar series "Wednesday Nights of Petri Nets and Their Extensions". Thus, for example, Fahland's synchronous proclets allow to describe the communication of multiple workflows, van der Aalst's object-centric nets can be discovered from logs where events have multiple case identifiers, whereas Rivkin and van der Werf propose formalisms to verify the interplay between processes and object-persistence models such as databases.

Inspired by the object-centric paradigm, we thus present in this talk a replay-based conformance method using conservative workflow colored nets. This model allows to describe the expected behavior of a system composed by end-to-end processes handling individual objects. Notably, when replaying an event log on top of our model, we consider a jump strategy where tokens representing objects move from their current location to input places of a transition to fire. Token jumps allow to unveil "desire lines", i.e., object paths unforeseen in the specification model. We introduce local conformance metrics based on the proportion of jumps in specific model components. The metrics allow to measure the magnitude of non-compliance in precise parts of a system. Supported by an implementation of our method, we present a practical example based on trading systems, where orders from users are matched to trade. Finally, we compare our approach with other well-known conformance methods.

WN-PNE from February till April 2021

Year 2021 has started… And it also seems that the COVID-19 pandemic will keep on raging all around the world for quite a bit. With that much of uncertainty, why not to spend lockdown evenings with a cup of warm tea and in the company of world-class researchers?  

Petri nets have been at the core of concurrency for more than 50 years. Many valuable, foundational properties were discovered for ordinary P/T-nets, in turn allowing them to be used for modelling and analysing complex systems. At the same time, ordinary P/T-nets are not expressive enough to attack concrete theoretical and industrial problems. This pushes both researchers and practitioners towards finding extensions that suit better their tasks.

WN-PNE is a virtual event that brings together Petri net researches and enthusiasts from all around the world working on the theory and practice of Petri nets and their extensions (data-aware, multi-agent, algebraic/logic, time etc.). The focus of the event is to present ongoing research as well as more foundational contributions to the domain, and provide opportunities for discussion on possible future developments. We also hope that WN-PNE will inspire young researchers and help to foster new collaborations.

https://pais.hse.ru/en/seminar-pne/ 

16.12.2020 On the membership problem for finite automata over symmetric groups

Speaker: Arthur Khashaev, Moscow State University.

We consider automata in which transitions are labeled with arbitrary permutations. The language of such an automaton consists of composition of permutations on all possible accepting computation paths. The membership problem for finite automata over symmetric group is the decision problem, where for a given automaton and a permutation it is asked whether the permutation belongs to the language of the automaton. It is shown that this problem is NP-complete, regardless of whether the automaton is acyclic. We also propose an efficient algorithm for the case of strongly connected automata.

03.11-04.11.2020 XI Workshop "Program Semantics, Specification and Verification: Theory and Applications" PSSV-2020

20.10.2020 Robustness measure for active management of collective investments

Speaker: Aliya Sharipova, PAIS Lab Research Fellow.

A practical approach to estimating of the investment strategy robustness is considered. As a quantitative measure of robustness, the objective

function smoothness degree and measure of stability of a strategy are proposed for utilization. After the optimization has been conducted, it is essential to utilize an additional criterion for the selection of strategies that possess better robustness property. The utilization of the quantitative estimate of the strategy robustness enables a better strategy to be chosen in the efficiency analysis of collective investments management.