Reading time:

Neural networks are powerful and flexible models that work well for many difficult learning tasks in image, speech and natural language understanding. However, designing these networks is difficult and time intensive. Can neural networks be used to design other neural network architectures quickly and efficiently? In "Neural Architecture Search with Reinforcement Learning" (goo.gl/drVqFT), Google Brain Resident (g.co/brain) Barret Zoph and Research Scientist Quoc Le explore the possibility.