Rodrigo Cesar Bonini

rodrigo.bonini@ufabc.edu.br

PhD Student
University Federal of ABC
Computer Science Department
Santo André, Brazil.

Lattes Profile Lattes Profile





Summary       Current Research Projects       Past Research Projects       Publications       Other Activities/Interests      



Summary


I received the Bachelor Degree in Computer Science from Pontifical Catholic University of São Paulo in 2013 and Post-graduate in Software Engineering at Pontifical Catholic University of São Paulo in 2015.

Meanwhile, I’ve also worked as Software Engineer and System Analyst Researcher at Itau Unibanco(the largest bank in Brazil), actively participating in many research and projects until 2015. After that, I received my Msc Degree in Computer Engineering in University of São Paulo in 2018, where I worked with Reinforcement Learning, Transfer Learning, Evolutionary Algorithms and some other Machine Learning techniques. During the Msc, I’ve worked full-time as an academic researcher, having won a group scholarship CEST Group(Centro de Estudos Sociedade e Tecnologia - Centre for Society and Technology Studies) award sponsored by Microsoft and contributing in discussions and work of the respective group. All these activities aforementioned provided me some experiences as proposing academic solutions to companies business and data problems, teaching assistant to undergrad subjects, and mainly to think about studying new solutions and abstractions in our society.

Currently, I'm coursing a PhD in Computer Science at the University Federal of ABC. My main goal now is to find useful applications of Reinforcement Learning, mainly in Biology, in order to help finding relationship between genes and try to control and intervene in Gene Regulatory Networks, consequently helping to find the cure for some diseases.

My research interests include Machine Learning, Reinforcement Learning, Transfer Learning, Systems Biology, Data Mining, Computer Vision, Smart Cities, Evolutionary Algorithms and Games .



Current Research Projects


Reinforcement Learning Applied to Control and Intervere in Gene Regulatory Networks (2019 -)

This research aims at try to model and test Reinforcement Learning Algorithms to allow knowledge generalization and reuse across similar but different tasks. With that, we expect to better control and infer in gene regulatory networks. The main objective of this research is to develop RL algorithms to successfully learn how to infer and intervene in GRNs to maintain a biological system in a desirable state. Solving this challenge includes (i) finding the best way of modeling the problem as an MDP; (ii) coping with the curse of dimensionality inherent from the domain; and (iii) proposing ways to safely evaluate the proposed methods without harming experiment subjects.

Machine Learning and Data Mining Applied to Decision Making Systems (2014-2015, 2019 -)

The use of data mining is already a reality in many organizations. However, previous studies use it without regard to users and customers needs regarding their priorities for their problems. Thus, this research conducts an analysis of data mining and machine learning techniques applied in a variety of scenarios, such as personnel selection process, forecasting, classification and frauds prevent, in order to discuss the influence of choosing these techniques and their performance evaluation metrics.

Learning Options for Transfer Learning in Multiobjective Reinforcement Learning Systems (2016 -)

Reinforcement Learning (RL) is a successful technique to train autonomous agents. However, the classical RL methods take a long time to learn how to solve tasks. Option-based solutions can be used to accelerate learning and transfer learned behaviours across tasks by encapsulating a partial policy into an action. However, the literature report only single-agent and single-objective Option-based methods, but many RL tasks, especially real-world problems, are better described through multiple objectives. Thus, this research aims at analyze methods for learn options in Multiobjective Reinforcement Learning domains in order to accelerate learning and reuse knowledge across tasks.

CEST Group Member(2016 -)

The Information and Communication Technologies have affected people’s way of life and created new habits and amenities that have been solving many of their everyday problems; however, they have brought together new problems such as those related to people’s safety, business, privacy, intellectual property, etc. The Society and Technology Study Center (or, CEST – Centro de Estudos Sociedade e Tecnologia, in Portuguese) was created at the Universidade de São Paulo in December, 2013, from grants by Microsoft Co. with the purpose of discussing these complex issues and providing meetings, involving developers, investigators, the society, and the government so as to organize these and other major issues and generate adequate answers and proposals, paving the way for solutions that go through habits, rules and developed applications. Some concepts transversal to all these themes such as the application into online education, creative currencies, Internet of Things (IoT) etc. are all themes prone to be problems or solutions of wide range and complexity, besides being the society’s general interest.


Evolutionary Adaptive Systems to Support Decision Making (2016 -)

More and more people need to consider a wide range of information for their decision- making, turning the society more dependent on computer systems that provide assistance. In particular, with respect to effectiveness and efficiency, some services can respond quickly to queries related to a certain region of space surrounding the user or, in other words, location-based systems. Currently, there are also initiatives to combine several attributes of the consultations,and often conflict with each other, with the use of multi-criteria and evolutionary algorithms based systems for decision making. Other initiatives seek to provide customised responses, where user preferences can be considered. However, few systems offer the service of responding these queries combining location-based systems with multi-criteria systems or evolutionary algorithms for decision-making and, even rarer, provide a customised response. This project aims not only at aggregating these three technologies - location-based systems, multi-criteria systems for decision making and systems that use the preferences of the consultant - in a system of decision support, but also to combine the power of machine learning, particularly reinforcement learning, to learn user preferences from the observation of their behaviour and choices in previous decisions.




Past Research Projects
Main Past Projects

Automatic Payment of Fees using Machine Learning (2012)

As Itau Unibanco is a large company, there were approximately 9000 new legal process entering every month, belonging to several different locations. The lawyers responsible for these processes used to receive their fees manually, and because of this large demand, they oftentimes received their fees late or with errors. Hence, my team proposed to use some machine learning techniques in order that the legal systems could pay the fees to the lawyers automatically.

Enmac: Application of a DSL as a tool to support programming education (2012-2013)

Enmac is a DSL that uses pseudocode and give some tips to students and beginners at programming. This tool was used in Pontifical Catholic University in some initial subjects of the courses of Information Systems and Computer Science to support the students. I had an active participation in the requirements elicitation, development and analysis of results in this project, also working as monitor of undergrad subjects, helping the students using the tool.

Qualified Registering of Legal Proceedings using OCR (2012 - 2013)

Optical character recognition(OCR) is the mechanical or electronic conversion of images of typed, handwritten or printed text into machine-encoded text, whether from a scanned document, a photo of a document, a scene-photo (for example the text on signs and billboards in a landscape photo) or from subtitle text superimposed on an image (for example from a television broadcast). As Itau Unibanco is a large company, there were approximately 9000 new legal process entering every month, and the department responsible for classify and designate these processes was unable to hang such demand. Hence, my team proposed to use OCR technologies to identify and classify the projects through some Pattern Recognition and Computer Vision techniques, alleviating the employees work load.

Machine Learning and Data Mining Application in Development Banking Systems (2014 - 2015)

Historically, banks contains lots of information about transactions, applications, data and investment of shareholders and customers. But, part of this information ends up not being used for getting lost amid massive databases of institutions and because of conventional programming techniques fail to be efficient to handle as much data. Thenceforth, this project aimed at investigate and find gaps and failures in banks business processes and systems. This study allowed me to propose some improvement in several processes, making these institutions significantly reduce their money spent.

Master Data Management of Products (2014 - 2015)

System evolution project of a small database of project for a Product MDM, based on a roadmap of improvements up from comparisons between marketing tools and internal tool institution at Itau Unibanco. I had active participation in the requirements gathering, analysis and definitions of solutions, also being responsible for communication with client area and the contracted suppliers for the project. I’ve also proposed and sketched the system with SOA architecture, having designed several web services and mainframe subprograms, facilitating its integration with another systems.

Master Data Management of Domains (2014 - 2015)

Project creation of a MDM (Master Data Management) domain from the mapping enterprise metadata and integration between the areas of Data Governance and Information Origin System at Itau Unibanco. I had active participation in the requirements gathering, analysis and definitions of solutions, also responsible for the communication with the client area and the contracted suppliers for the project. During this period, I’ve also been responsible for an analysis of queries with bad performance in several systems, modelling solutions and implementing them.



List of Past Publications


Learning Options for Transfer Learning in Multiobjective Reinforcement (WPGEC, São Paulo, 2016)

Learning Options for Transfer Learning in Multiobjective Reinforcement (AAAI, San Francisco, 2017)

Using Options to Accelerate Learning of New Tasks According to Human Preferences (AAAI Human Machine Collaborative Learning Workshop, San Francisco, 2017)

Transferring Probabilistic Options in Reinforcement Learning (AAMAS Workshop on Transfer in Reinforcement Learning, São Paulo 2017)

Transferring Probabilistic Options in Reinforcement Learning (WPGEC, São Paulo 2017)

A Framework to Discover and Reuse Object-Oriented Options in Reinforcement Learning (BRACIS, São Paulo 2018)



List of Current Working on Publications


Speeding up Reinforcement Learning for Inference and Control of Gene Regulatory Networks (in progress, expected to be published in AAAI, New York 2020)

An analysis of data mining approaches applied in a personnel selection process (in progress, expected to be published in AAAI, New York 2020)

Reinforcement Learning for Inference and Control of Gene Regulatory Networks (in progress, expected to be published in Workshop Latinx at Neurips, Vancouver 2019)

Speeding up Multiobjective Reinforcement Learning Using Single-Objective Options Discovered in Different Tasks (in progress, expected to be published in AAMAS, Auckland 2020)

A Framework to Learn and Reuse Options in Multiobjective Reinforcement Learning (expected to be published in 2020)

A Study of Different Types of Predictors (expected to be published in 2020)

Feature Selection for Psychometric Data in Human Resources Selection Processes (expected to be published in AAAI 2020 or Neurips 2019 Workshops)




List of Honors and Awards


Programming Marathon/Competition (PUCSP, São Paulo, 2010)

Enmac: Application of a DSL as a tool to support programming education - Distinguished Undergraduate Thesis (PUCSP, São Paulo, 2012)

Multicriteria Adaptive Systems to Support Decision Making - CEST/Microsoft Group (Centro de Estudos Sociedade e Tecnologia) Scolarship/Award (USP, São Paulo, 2016-2017-2018)

Learning Options for Transfer Learning in Multiobjective Reinforcement Learning - Distinguished Paper Award (Workshop de Pós-Graduação - Engenharia da Computação (WPGEC), São Paulo, 2016)

Learning Options for Transfer Learning in Multiobjective Reinforcement Learning - Best Master Degree Student Poster Award (Workshop de Pós-Graduação - Engenharia da Computação (WPGEC), São Paulo, 2016)

Learning Options for Transfer Learning in Multiobjective Reinforcement Learning, Using Options to Accelerate Learning of New Tasks According to Human Preferences - Student Travel Grants (AAAI, San Francisco, 2017)

Transferring Probabilistic Options in Reinforcement Learning - Student Travel Grants (AAMAS, São Paulo, 2017)

Transferring Probabilistic Options in Reinforcement Learning - Best Master Degree Student Poster Award (Workshop de Pós-Graduação - Engenharia da Computação (WPGEC), São Paulo, 2017)

A Framework to Discover and Reuse Object-Oriented Options in Reinforcement Learning - Distinguished Paper Award (BRACIS, São Paulo, 2018)


Other Activities/Interests

Music - I’m trying to play electric guitar since I was 14 years old. When I was a teenager, I dreamed of being a great and rich musician to be able to travel around the world and be recognized by everyone. Unfortunately, I’m far from the professional level because it takes too much time and dedication. Nevertheless, I believe it's never too late to dream and that we should never give up, so I’m working hard on it until nowadays.

Games - Since I know myself as somebody, I play video-games. Everything started when I played Mortal Kombat and Sonic the Hedgehog of Megadrive for the first time. After that, it’s hard for me to spend some time without playing or having fun with a game.

Films/TV Series - In my free time, I like a lot watching films, animes and series, specially the action and fiction ones.

Languages/Travels - I’ve interest in learning new languages and studying different cultures and behaviour. Although I can speak English, Spanish and a little of German and Italian, it’s so hard to practice some new language in Brazil, since it’s a large country and most of the people speak only Portuguese. So, I want to travel around the world seeking to learn different things about the world and the people.