Ricardo Cerri, a computer scientist with a lifelong fascination for biology, completed his undergraduate studies at the Universidade Estadual Paulista Júlio de Mesquita Filho. He further pursued his master’s and doctoral degrees in computer science and computational mathematics at the University of São Paulo. During his doctoral studies, he conducted part of his research at the Universities of Surrey and Kent in the UK. In 2014, his thesis was recognized as the second best in a Brazilian competition for theses on artificial intelligence organized by the Brazilian Computer Society. He completed his post-doctoral studies at the University of Porto in Portugal. Cerri’s work involves developing algorithms inspired by the theory of evolution that can program neural networks autonomously.
A music enthusiast, Cerri explores the diverse facets of this art form. He plays the double bass and guitar in a band with friends and has been a part of choral groups since his undergraduate days at Unesp. He is also a singer and never misses an opportunity to dance to forró. Currently, he is a lecturer in the Department of Computer Science at the Federal University of São Carlos.
Artificial Neural Networks (ANNs) are computer models that draw inspiration from the brain, where interconnected artificial neurons execute tasks such as image classification. Deep Neural Networks (DNNs), characterized by numerous layers of neurons, represent the cutting edge in tackling complex problems. While they can be manually designed, this requires specialized knowledge. Consequently, researchers have been exploring the automatic construction of ANNs, a field known as neuroevolution. This raises the question: Can DNNs be built automatically, without manual intervention, and still effectively solve problems? To address this, we will employ evolutionary algorithms. Inspired by the theory of evolution, these algorithms encode DNNs as individuals in a population, seeking the most effective ones for problem-solving. Given the diversity of individuals (DNNs) within a population, we aim to automatically discover DNNs with novel architectures and configurations that are not yet documented in the literature.