Esteban Real, Software Engineer, Google Brain
Evolving Machine Learning Algorithms
The effort devoted to hand-crafting machine learning (ML) models has motivated the use of automated methods. These methods, collectively known as AutoML, can today optimize the models' architectures to surpass the performance of manual designs. I will discuss how evolutionary techniques can allow AutoML not only to perform such optimization, but also to discover complete ML algorithms from scratch. Using only basic mathematical operations as building blocks, our experiments give rise to ML techniques such as backpropagation, simple neural networks, and weight averaging. This is the case even if derivatives are not provided among the building blocks: gradients simply arise as a consequence of the search process. I will pay special attention to the role of death in allowing evolution to handle noisy measurements.
Bio: Esteban Real is a staff member of the Brain team in Google Research, where he's been since 2015. He completed his Ph.D. in Physics at Harvard University, where he worked on computer models that learn from living neural networks. Since then, he has been interested in biologically inspired computation. His current work focuses on the relationship between evolutionary algorithms and automated machine learning (AutoML), a field in which computers must discover how to learn.