13/02/2020

Viana: 'Mathematics has enormous potential to be explored'

How much are mathematical sciences worth? There is still no data available capable of measuring the impact of mathematics on the Brazilian economy, “but there is enormous potential to be explored in this regard. And Brazil has the necessary expertise ,” said the director-general of IMPA, Marcelo Viana. At the opening of the 1st Mathematics and Industry Workshop , this Thursday (13), Viana presented research done in other countries that calculated how much mathematics adds to the economy.

“A study conducted in the United Kingdom in 2012 showed that the sector generated 1.8 million jobs, which corresponds to 10% of the workforce. The productivity of these roles is double the national average. We knew that mathematics was important, but we had no idea it was this important,” Viana highlighted.

Foto: Marcelo Viana, diretor-geral do IMPA

Organized in partnership with the Center for Mathematical Sciences Applied to Industry (CeMEAI) at USP, the workshop aims to bring the productive sector closer to academia. Over the two days of programming, more than 300 students, researchers, and business managers will learn about mathematical tools for solving real-world problems.

“Artificial intelligence has existed for many years. But why this 'new wave' now? The most frequent use of artificial intelligence is due to the interaction between mathematics, which provides the background , and computing,” highlighted José Alberto Cuminato, director of CeMEAI.

Foto: José Alberto Cuminato, diretor do CeMEAI

In his presentation, Cuminato introduced the institution and its potential to contribute to the industry, reinforcing the importance of a fundamental relationship. "Mathematics and business combine, and yield good results."

Alfredo Renault, Superintendent of Research and Technological Development at the National Petroleum Agency (ANP), believes in the success of the partnership between mathematics and oil companies. “It was through the ability to solve the descriptions that physics made of phenomena that we were able to do what is most important within this context of knowledge, which is to make predictions. To predict when a building will be completed, to predict that oil will flow from the well in a certain way. Mathematics allows us to make predictions and for things to happen more independently of the empirical process,” Renault emphasized.

Foto: Alfredo Renault, superintendente de Pesquisa e Desenvolvimento Tecnológico da ANP

For Renault, the Mathematics and Industry Workshop demonstrates IMPA's potential to contribute to companies. "The oil industry is moving towards a third stage, which is digital transformation, becoming possible as we expand our data processing capacity. And IMPA certainly has the necessary elements to offer to our productive sector."

Researchers presented applications of mathematics.

With clear presentations, researchers from IMPA and USP showed examples of the application of mathematics in artificial intelligence, machine learning , data science, and finance.

Luis Gustavo Nonato, from the Institute of Mathematical and Computer Sciences at the University of São Paulo (ICMC-USP), presented a project he has been developing since 2016, in which he applies data science to security policies. In partnership with the Government of the State of São Paulo, the Municipality of São Carlos, and higher education institutions, the researcher has already developed, along with collaborators, three tools to analyze the pattern and incidence of robberies in areas of the city of São Paulo.

“An initial difficulty was gaining the trust of the government departments. As we showed results, they offered us more information, and that's how we were able to build an even larger database,” he said. Through the analysis of temporal patterns of crimes, the group was able to identify the 1,500 most dangerous points in the city among the more than 19,000 points where robberies had been reported.

Machine learning and probability were the topics addressed by IMPA researchers Roberto Imbuzeiro and Paulo Orenstein. “What Rio de Janeiro resident isn't used to hearing the phrase 'in a single day it rained the equivalent of ⅓ of what was predicted for the entire month'?” asked the institute's researcher Roberto Imbuzeiro. If the forecast is frequently exceeded, it's because “rainfall is a difficult phenomenon to predict,” stated Imbuzeiro.

The mathematician warned of the need for a deeper understanding of theoretical areas of mathematics for a greater comprehension of tools such as machine learning . “Neural networks involve many parameters and are very vulnerable to attacks. Currently, the coronavirus counting network is a focus of interest for researchers in the field. But understanding it is very costly.”

IMPA researcher Alexei Mailybaev presented two doctoral projects from the Fluid Dynamics Laboratory, linked to artificial intelligence and computational challenges. One approach used by the research group includes the application of machine learning methods to model turbulent flows. “There are very advanced computational programs, but they cannot access the structure at the level we are doing. I believe our simulator is well ahead of the others.”

The mathematical economics area at IMPA was presented by researcher Luciano de Castro. “It’s a very strong niche at IMPA, but few people know it exists. The group of researchers and students is small, but they have a very strong connection with the rest of the people at the institute, which is enriching,” he commented. Luciano presented the main themes of the area, such as general equilibrium, decision theory, information asymmetry, and game theory. He also spoke about new frontiers that are gaining importance, such as market design and financial innovations.