Alexandre Lemos is a Senior Research Scientist at Outsystem. He just delivered his Ph.D. thesis at IST. The topic of his Ph.D. thesis is Solving Scheduling Problems under disruptions. The work developed has placed among the five finalists of the International Timetabling Competition 2019.
During his Ph.D. he was a teaching assistant at the Department of Computer Science and Engineering of IST.
For his master thesis, Alexandre worked on Repairing Biological Networks using ASP and SAT based tools.
PhD in Information Systems and Computer, 2016-2021
Instituto Superior Técnico
MSc in Information Systems and Computer, 2014-2016
Instituto Superior Técnico
BSc in Information Systems and Computer, 2011-2014
Instituto Superior Técnico
From 25 Aug - Fri 28 Aug, PATAT 2020
From May 26 to May 29, CPAIOR 2020
From Jan 13 to Jan 18, 2019 Data61 International Optimisation Summer School
From Jul 7 to Jul 12, SAT 2019
From Jul 2 to Jul 5, EAIA 2018 - DS4BD Advanced School on Data Science for Big Data
From Jun 22 to Jun 25, SAT/SMT/AR Summer School 2016
From Set 8 to Set 9, INFORUM Simpósio de Informática 2016
2019 Mai 23, Dia do Tecnico 2019
From 2019 Apr 9 to Apr 10, PhD Open Days 2019
From Jan 27 to Jan 28, NDA
The tool developed to solve ITC-2019 instances. The tool is in the top 5 of the competition.
Boolean logical models of biological regulatory and signalling networks are increasingly used to formally describe and understand complex biological processes. Such models often need to be repaired whenever new observations become available, rendering the model inconsistent with the new observations. Although many studies have focuses on inferring new models (or classes of models) from data, the automation of model repair is still in its infancy, often still being a manual process and therefore prone to errors. In this work, a tool is provided to suggest possible repair operations to inconsistent models based on changing the model logical functions. The tool uses Answer Set Programming to identify possible repair operations to the models following both synchronous or asynchronous updating schemes. The tool was validated using both Boolean models of randomly generated networks and data sets of Escherichia coli at steady state. The limitations of the implemented tool for each of the updating schemes are discussed. The impact of the minimization algorithm is also addressed.
This paper discusses the problem of room usage usage optimization for university timetables. Given a timetable, the goal is to optimize the room occupation by changing the events allocated to each room. The methods proposed do not solve the problem from scratch and so no curriculum constraints are handled. The main objective is to optimize the timetable scheduled for each room without changing the already defined time slots for each lecture. This type of optimization is important since room space is a precious commodity. The proposed algorithms optimize the space allocation in terms of capacity (seating students) and compactness (reducing the number of gaps in the timetable for each room). This problem is analysed and discussed for the case study of Instituto Superior Técnico (IST), the engineering school from Universidade de Lisboa. At IST, the generation of timetables is still hand-made and therefore prone to optimization. This paper proposes a two-stage Integer Linear Programming (ILP) method to optimize the room usage. In the first stage, the goal is to assign lectures to rooms respecting their capacity. The second stage minimizes the gaps by reassigning the lectures to rooms (ensuring the same quality in terms of seated students found in the first stage). This implementation is able, in one hand, to prove optimality in both stages; on the other hand, it is quite time consuming. To address this issue, in this paper we also propose a greedy algorithm. The best cost function used to guide the algorithm is proved to be monotone, positive and sub-modular. Therefore, the results obtained are guaranteed to be within 63 percente of the optimal. The proposed solutions are tested using data from the lectures of Instituto Superior Técnico taught on the Taguspark and Alameda campus in the academic year of 2016⁄2017. The greedy algorithm is two orders of magnitude faster than ILP when considering large data sets. In terms of seated students, the best cost function is on average 2 percente below the optimal value. The timetable for a given room can be evaluated in terms of compactness (the number of transitions from free to occupied). Considering this metric, the solution provided by a greedy algorithm is only 1.2x worse than the optimal value for the Taguspark data sets. For the Alameda data sets, the solution obtained by ILP implementation is 0.7x better (optimal solution) than the greedy algorithm. However, the ILP implementation is, in the worst case, 5 orders of magnitude slower. Overall, both implementations provide significant gains for both optimization criteria when compared with the current hand-made approaches.
Models of biological regulatory networks are increasingly used to formally describe and understand complex biological processes. Such models are often repaired whenever new observations become available, because the model cannot generate behaviours consistent with the new observations, or because the behaviours are contradictory. This process of model repair is often manual and therefore prone to errors. In this work, we describe biological regulatory networks using the Boolean logical formalism, where nodes are represented by Boolean variables denoting biological components and edges denote regulatory interactions between components. The evolution of each variable is defined by a Boolean logical function depending on the values of the regulators of the component. Here, we propose to repair the model by changing inconsistent functions, with four types of atomic repairs which can be further combined. The goal is to find the cardinality minimal set of repairs allowing the model to satisfy all available observations.
Interactive Bar Table implemented for the Machine-Human interaction course. The project was made in collaboration with Cristina Verdasca and Pedro Moniz
A secure mobile payment system implemented in Android for the Network and Computer Security course. his work was developed in collaboration with João Luís e João Carraquico.
A simplified implementation of the Map Reduce middleware and programming model in C# for the Design and Implementation of Distributed Applications course. his work was developed in collaboration with João Luís e João Carraquico.