Program

Morning8:00 – 9:00Registration
9:00 – 9:30Welcome & Open Ceremony
9:30 – 10:30Fernando Buarque – Responsible AI, a primer (due to health related travel problems, the talk by Sanaz Mostaghim was cancelled)
10:30 – 11:00Coffee Break
11:00 – 12:00Paper Session In Person 1 (3 papers)
Lunch12:00 – 13:30Time for Lunch
Afternoon13:30 – 14:30Paper Session Remote 1 (3 papers)
14:30 – 15:30Rosangela Ballini – Adaptive Fuzzy Systems in Economics and Finance: Evaluating Interval Forecasts of High-Frequency Data
15:30 – 16:00Coffee Break
16:00 – 17:00Paper Session In Person 2 (3 papers)
17:00 – 18:00Paper Session In Person 3 (3 papers)
18:00Opening Ceremony
Morning8:00 – 8:30Registration
8:30 – 9:30Leslie Pérez Cáceres – Automatic Algorithm Configuration: Performance, Analysis and Design
9:30 – 10:30Paper Session In Person 4 (3 papers)
10:30 – 11:00Coffee Break
11:00 – 12:00Paper Session Remote 2 (3 papers)
Lunch12:00 – 13:30Time for Lunch
Afternoon13:30 – 14:30Gabriela Ochoa – Neuroevolution Landscapes and Trajectories
14:30 – 15:30Industrial Session
15:30 – 16:00Coffee Break
16:00 – 17:00Thesis Symposium
17:00 – 18:00Paper Session Remote 3 (3 papers)
20:00Gala Dinner
Morning8:00 – 8:30Registration
8:30 – 9:30Gerardo Rubino – Using Machine Learning in communication network research
9:30 – 10:30Paper Session In Person 5 (3 papers)
10:30 – 11:00Coffee Break
11:00 – 12:00Paper Session In Person 6 (3 papers)
Lunch12:00 – 13:30Time for Lunch
Afternoon13:30 – 14:30Paper Session Remote 4 (3 papers)
14:30 – 15:30Paper Session Remote 5 (3 papers)
15:30 – 16:00Coffee Break
16:00Awards and closing ceremony

Track Contents

In Person Session #1

11:00 – 11:20794Combining optimization and fire simulation modeling to protect biodiversity values at landscape scaleRodrigo Mahaluf, Jaime Carrasco, Fulgencio Lisón, Cristobal Pais, Alejandro Miranda, Felipe de la Barra and Andrés Weintraub
11:20 – 11:402939Chatbot use for pre-triage procedures: a case study at a free-service university dental clinicDouglas Vidal, Luana Pantoja, Fernanda Jassé, Diandra Arantes and Marcos Seruffo
11:40 – 12:009759Finding Frequent Patterns in a Technological Education Program of Pernambuco, BrazilEmilia Rahnemay Kohlman Rabbani, Juliana de Souza Rebouças, Márcia Maria de Albuquerque Neves, Gabriel Magalhães da Luz, Willian Vieira Do Nascimento, Gustavo Henrique Magalhães da Luz, Felipe Guerra Lago, Maria Celeste de Sousa Maia, Maicon H. L. Ferreira da Silva Barros, Patricia Takako Endo and Carmelo José Albanez Bastos Filho

Remote Session #1

13:30 – 13:50242The Influence of Different Text Structuring Methods on the Clustering of Game ReviewsAnara Olimpio, Pollyana Notargiacomo and Leandro de Castro
13:50 – 14:101117Ensemble Learning Models Applied in Energy Time Series of a University BuildingIsaakc Ortiz-Aguirre, Mayken Espinoza-Andaluz and Julio Barzola-Monteses
14:10 – 14:307430Evaluation of Multi-objective Evolutionary Algorithms for the Design of Resilient OTN over DWDM NetworksArysson Oliveira, José Cleyton Silva, Danilo Araújo, Joaquim Martins Filho and Carmelo J. A. Bastos Filho

In Person Session #2

16:00 – 16:202167Kaizen Programming for predicting numerical linear algebra operations performanceJimena Ferreira, Ernesto Dufrechou and Martín Pedemonte
16:20 – 16:408095A Many-Objective Optimization Approach to Generate Synthetic Datasets based on Real-World Classification ProblemsSteffano X. Pereira, Pericles Miranda, Thiago França, Carmelo J. A. Bastos Filho and Tapas Si
16:40 – 17:002227Application of a Genetic Algorithm with Social Interaction to Search Based TestingAdilson Neto, Rodrigo Pereira and Roberto Oliveira

In Person Session #3

17:00 – 17:20153Early prediction of generalized infection in intensive care units from clinical data: a committee-based machine learning approachFlavio Fonseca, Arianne S. Torcate, Ana Clara G. da Silva, Victor W. Freire, Gilles P. M. de Farias, João F. L. de Oliveira, Flávio M. de Oliveira Júnior, José Carlos da Silva Júnior, Dayane A. Gomes and Wellington Pinheiro Dos Santos
17:20 – 17:403889A comparison of Generative Adversarial Networks for image super-resolutionPatricia Cobelli, Sergio Nesmachnow and Jamal Toutouh
17:40 – 18:009125Hyperparameter Optimization for Convolutional Neural Networks with Genetic Algorithms and Bayesian OptimizationDavid Puentes, Carlos J. Barrios H. and Philippe O. A. Navaux

In Person Session #4

09:30 – 09:503052Generation of English Question Answer Exercises from Texts using Transformers based ModelsGonzalo Berger, Tatiana Rischewski, Luis Chiruzzo and Aiala Rosá
09:50 – 10:108640Autoencoder with adaptive and trainable activation function to compress imagesJoão Henrique de Medeiros Delgado and Tiago Alessandro Espinola Ferreira
10:10 – 10:309829Classification of patients with diabetes by analyzing the patterns of proteins present in saliva using machine learningRodrigo Costa, Carmelo J. A. Bastos Filho and Anthony Lins

Remote Session #2

11:00 – 11:204751Boolean Binary Grey Wolf OptimizerRodrigo Cesar Lira Da Silva, Mariana Gomes Motta Macedo, Hugo Valadares Siqueira and Carmelo José Albanez Bastos-Filho
11:20 – 11:405917Assignment of bug reports to software developers using a multi-population evolutionary methodKannya Araujo, Luiz Mendes, Guilherme Avelino, Ricardo Rabelo and Eneko Osaba
11:40 – 12:009714Tuning of PI Controllers for Electric Drives using Evolutionary Multi-Objective Optimization AlgorithmGuilherme Fernandes dos Santos, Wander Gonçalves da Silva and Gélson da Cruz Júnior

Industrial Session

14:30 – 14:50Audio processing for diarization and its application to the analysis of primary school courses – Braulio Ríos – Ceibal
14:50 – 15:10Machine learning applied to digital and traditional commerce – XmartLabs
15:10 – 15:30Inteligencia artificial aplicada a Infraestructura Urbana (in Spanish) – Daniel Muniz – Intendencia de Montevideo

Audio processing for diarization and its application to the analysis of primary school courses

The problem of speaker diarization (or just diarization) consists of answering the question «who spoke when?», taking as input an audio signal, which typically has many different speaker voices present on it. In this talk we will briefly introduce the diarization problem and its applications, covering some of the state of the art audio processing algorithms used to tackle it. Finally, we will present the work in progress that is being carried out within a research project between Ceibal and the audio processing group of the Facultad de Ingeniería, UdelaR. In this case, the aim is to adapt the techniques used for diarization to the particular case of teaching analysis in primary school courses. The goal is to develop tools that assist the education technicians who carry out the evaluation and monitoring of different Ceibal educational programs.

Braulio Ríos is an MSc Student in Data Science and Machine Learning at UdelaR, currently working on his thesis about audio processing for diarization and classification, in the context of a research project in collaboration with Ceibal. He has experience working as a developer and consultant for the software industry, mainly in R&D projects with Machine Learning components for the past five years.

Machine learning applied to digital and traditional commerce

We’ll present three case studies illustrating how machine learning can be applied to digital and traditional commerce solutions.

Inteligencia artificial aplicada a Infraestructura Urbana (in Spanish)

Se presenta una prueba piloto, que tiene como objetivo analizar la viabilidad de detectar a través de Analítica de video ciertos aspectos urbanísticos de la Ciudad. Se trabajó en reconocimientos de patrones utilizando analíticas de video, el piloto fue realizado por la Intendencia de Montevideo en conjunto con la empresa Digital Sense. Se presenta una descripción de los tipos de pilotos que se hicieron, con que tecnologías se trabajó, sobre que aspectos de la infraestructura Urbana se realizaron y también se presentan resultados (porcentajes de detección). Se trabajó en detección de: Rajaduras de Pavimento y Estado de la cartelería (señalización vertical).

Daniel Muniz es Ingeniero en Computación egresado de UDELAR. Entre los años 2000 y 2012, trabajo como desarrollador y también como líder técnico en el Sistema de Transporte Metropolitano de la Intendencia de Montevideo. Luego participó en varios proyectos de tecnología, asesorando licitaciones y compras, y desde 2016 es el coordinador técnico del Centro de Gestión de Movilidad representando al Departamento de Desarrollo Sostenible e Inteligente.

Thesis Symposium

16:00 – 16:20MapView: Exploring Datasets via Unsupervised View Recommendation – Master’s thesis – Thiego Buenos Aires de Carvalho – University of Pernambuco
16:20 – 16:40Novel Data Mining Methods and Applications in Combinatorial Optimization and Bioinformatics – Ph.D. thesis – Marcelo Rodrigues de Holanda Maia – Universidade Federal Fluminense
16:40 – 17:00Automatic Algorithm Configuration: Methods and Applications – Ph.D. thesis – Marcelo de Souza – Universidade Federal do Rio Grande do Sul

Remote Session #3

17:00 – 17:204142A Transfer Learning Approach for the Tattoo Classification ProblemRodrigo Tchalski da Silva and Heitor Silvério Lopes
17:20 – 17:407730Student Dropout Prediction using 1D CNN-LSTM with Variational Autoencoder OversamplingEduarda Coppo, Rhuan Caetano, Leandro de Lima and Renato Krohling
17:40 – 18:009649Classification of Legal Documents in Portuguese Language Based on SummarizationMarie Chantelle Cruz Medina, Lucas Matheus da Silva Oliveira, Jean Felipe Coelho Ferreira, Leandro Honorato de S. et al.

In Person Session #5

9:30 – 9:509426A Legal Information System for Intelligent Sentence Mining Applied to Civil LawDavid Barrientos, Bruno Fernandes, Cleyton Rodrigues, Leandro Silva, Allana Rocha, Paulo Sobral, Bruno Souza, Dionizio Feitosa, Juliana Barreto and Mabel Guimaraes
9:50 – 10:108928Applying non-compensatory multicriteria methods to build Better Life Index countries rankingGlaucia da Costa Azevedo and Helder Gomes Costa
10:10 – 10:30Industrial Session: UTE talk

In Person Session #6

11:00 – 11:204116Scheduling of the Uruguayan Football and Basketball LeaguesMartín Lago, Nicolás Lantean, Santiago Rodríguez, Federico Defranco and Martín Pedemonte
11:20 – 11:404673Comparative Study of PID, DMC and Fuzzy PD+I Controllers in a Control Laboratory KitDiego Páez Ardila, Diana Martinez, Cesar Valencia, Ricardo Tanscheit and Marley Vellasco
11:40 – 12:006132A New evolving Fuzzy System with Mechanisms to Deal With Uncertainties in Times Series ForecastingKaike Alves, Direnc Pekaslan, Christian Wagner and Eduardo Aguiar

Remote Session #4

13:30 – 13:504760Self-explanatory error checking capability for classifier-based Decision Support SystemsFlávio Rosendo da Silva Oliveira and Fernando Buarque de Lima Neto
13:50 – 14:108353Passive Sonar Classification Using Time-Domain Information and Recurrent Neural NetworksMarlon Souza, Natanael Júnior and José Seixas
14:10 – 14:305863Clustering-and-Bagging-based Ensemble for Novelty Detection in Passive Sonar SystemsEduardo Sperle Honorato, João Baptista de Oliveira Souza Filho and Victor Hugo da Silva Muniz

Remote Session #5

14:30 – 14:509409Machine Learning Approach for Trend Prediction to Improve Returns on Brazilian Energy MarketMoisés Santos, Douglas Braz, André Carvalho, Renato Tinós, Marcos Paula, Gabriel Doretto, Ewerton Guarnier, Donato Silva Filho, Danilo Suiama, Lorena Ferreira and José Carmo Júnior
14:50 – 15:105144Hybrid Models Based on Error Correction to Predict Thermoelectric Power Plants Fuel ConsumptionElias Amancio Siqueira-Filho, Maira Farias Andrade Lira, Hugo Valadares Siqueira and Carmelo J. A. Bastos-Filho
15:10 – 15:309894Reinforcement Learning applied to the Routing and Spectrum Assignment in Elastic Optical NetworksSantiago Arce, Luis Ayala Albertini, Ivan Ríos, Diego Pedro Pinto Roa, José Colbes and Marcos Villagra

Keynote Speaker 1

Fernando Buarque

Responsible AI, A Primer

Current advances in Machine Learning and Artificial Intelligence are many and often surprising. So far, such advances have mostly focused on qualitative results (e.g., creating metaheuristic algorithms that can solve complex problems), which generally make users happy. However, the fact that most algorithms produced and used are oblivious of what they are actually performing, and possess little social accountability, somehow are massive shortcomings. In this key speech, I will share some important concepts and questions, as well as ideas and preliminary work being carried out in our research group. These, so that more reasonable, responsible and conscious type of AI could be yielded. None the less, results at this point are not many neither impressive, but the importance of the discussion is deemed to be paramount. Some input from the audience will be highly valued on this multidisciplinary and pressing theme.

Keynote Speaker (talk cancelled due to health-related travel problems)

Sanaz Mostaghim

Recent Advances in Swarm Intelligence and Swarm Robotics

In the past decades, we witnessed a large improvement of autonomous systems. Today, such systems are everywhere and enable us to handle complex problems in industrial and scientific applications. However, they also pose new challenges for the development of algorithms to design and control them. One challenge concerns the large amount of such systems which are able to communicate with each other and hence produce a large complex system. Looking at nature, biological systems solve complex tasks using decentralized and simple structures. In this talk, we aim to give an overview into such nature-inspired algorithms such swarm intelligence and describe their applications in autonomous systems. Swarm intelligence is a collective learning process which can lead to a self-organized system of simple individuals, which together create a global emergent behavior.

Such systems can adapt very well to changes in the environment and produce flexible and at the same time robust behaviour. One advanced application of swarm intelligence is in the area of swarm robotics in which simple small robots can collectively learn to achieve some predefined complex tasks. In this talk, the algorithms of swarm intelligence are presented, analyzed and compared. The following topics will be covered:

  • Fundamentals of swarm intelligence algorithms and optimization
  • Collective learning and decision-making
  • Collective perception algorithms
  • Control mechanisms for self-organized systems using the environment (isomorphic and non-isomorphic transformations)
  • Swarm and evolutionary robotics

Keynote Speaker 2

Rosangela Ballini

Adaptive Fuzzy Systems in Economics and Finance: Evaluating Interval Forecasts of High-Frequency Data

The forecast of the future movement of economic and financial variables assumes a central role for the composition of portfolios, risk management, asset pricing and investment analysis; therefore, the development of prediction methodologies is of fundamental importance. With the recent and rapid growth in the availability of financial information, especially at intraday frequencies, approaches to forecasting interval time series have gained prominence in the
literature, since they comprise the construction of more informative forecasts, capable of capturing the fluctuations of an asset, index or
rate over the course of a transaction day, as opposed to techniques based on one-off anticipations. Adaptive fuzzy models are nonlinear, are able to update their structure and functionality according to data streams, and handle uncertainty from fuzzy sets. Thus, this approach allows the dynamic treatment of complex phenomena, as well as considering information affected by uncertainties, as is the case of financial markets.

Keynote Speaker 3

Leslie Pérez Cáceres

Automatic Algorithm Configuration: Performance, Analysis and Design

Algorithm configuration is the task of finding a set of parameter settings which allow an algorithm to reach an expected performance level. All sorts of algorithms are designed to expose parameters for their later configuration. These postponed design choices grant algorithms flexibility and allow them to be general enough to tackle different problems. The algorithm configuration task is often challenging; finding adequate settings usually requires some level of expert knowledge. From a practical perspective, this task can be tedious and often computationally expensive. In this talk, we will discuss the problem of algorithm configuration, how configurators automate this task and how these techniques can be helpful in understanding algorithms. Automated configuration not only eases algorithm application but also provides a tool that can be used to support automated algorithm design. We will discuss what is required to achieve such a level of automation and the challenges associated with this goal.

Keynote Speaker 4

Gabriela Ochoa

Neuroevolution Landscapes and Trajectories

Neuroevolution, the use of evolutionary algorithms to design neural networks, has a long tradition in evolutionary computation with roots in the early 1990s. Traditionally, neuroevolution systems optimize both the neural network topology and its weights. However, when scaling up to contemporary deep models with millions of weights, gradient-based weight optimization generally outperforms evolutionary methods. In consequence, many recent methods evolve the topology only. This approach is also known as Neural Architecture Search (NAS). This talk overviews our recent work on modelling neuroevolution and NAS systems with search trajectory networks (STNs) and local optima networks (LONs) with the aim of providing a visual and quantitative understanding of search and optimisation in this domain.

Keynote Speaker 5

Gerardo Rubino

Using Machine Learning in communication network research

Nowadays, Machine Learning (ML) tools are commonly used in every area of science or technology. Networking is not an exception, and we find ML all over the research activities in most fields composing the domain. In this talk, we will briefly describe a set of research activities we have developed along several years around several pretty different families of problems, using ML methods. They concern:

  1. the automatic and accurate real time measure of the Quality of Experience of an application or service built on top of the Internet around the transport of video or audio content (e.g. video streaming, IP telephony, video-conferencing, etc.),
  2. network tomography (measuring on the edges to infer values inside the network),
  3. time series forecasting in several contexts, in particular concept drift detection or anomalies detection, and
  4. service placements in Software Defined Networks, a central problem in 5G and B5G technologies.

The corresponding ML tools are mainly Supervised Learning and Reinforcement Learning, even if we are currently using Unsupervised Learning in recent activities of point 1. After this global presentation we will make one or two zooms on some specific results we obtained with these powerful tools, and some of the current projects we are currently developing.

Organizers

IEEE Computer Intelligence Society
IEEE Computer Intelligence Society

https://cis.ieee.org/

IEEE
IEEE

https://www.ieee.org/

Instituto de Computación
Instituto de Computación

https://www.fing.edu.uy/inco

Facultad de Ingeniería
Facultad de Ingeniería

https://www.fing.edu.uy/

Universidad de la República
Universidad de la República

https://www.universidad.edu.uy