Research publications

Publications

You can also view my publications on Google Scholar.

Automated Architecture Design for Deep Neural Networks
Steven Abreu
ArXiv (link)

2019 Dean's Prize(Best Bachelor Thesis in Engineering, Computer Science, Mathematics)

Bachelor Thesis. Investigated techniques that automate the architecture design for deep neural networks (AutoML), both through neural architecture search and through dynamic learning (i.e. modifying the network structure while learning). Proposed a framework for evaluation. Proposed an extension for the Forward Thinking framework (Hettinger et al, 2017) for increased autonomy.

Research projects

Research Experience

November 2020 - present
Early Stage Researcher (PhD)
MINDS @ University of Groningen (link)
Groningen, Netherlands

Joined the MINDS research group at RUG as part of the European post-digital training network. The post-digital goal is to facilitate the interdisciplinary education and research needed for the design and development of unconventional computing systems. At MINDS, we try to understand how intelligent information processing can arise from dynamics in nonlinear material substrates. We are working toward a fundamental theory of computing in brainlike materials, grounded both in physics and mathematics.

More information: MINDS, Post-digital ETN.

September 2018 - December 2018
Research assistant
DELPHI @ Carnegie Mellon University (link)
Pittsburgh, USA
Worked on epidemiological forecasting using machine learning. Worked with two PhD students on expanding the existing nowcasting system for influenza to dengue and norovirus.

∆ steven abreu 2020