Vivienne Sze
Papers
22
Total Citations
4,576
H-Index
12
About
Vivienne Sze is an associate professor at MIT whose research sits at the intersection of energy-efficient hardware design, deep learning, and autonomous systems. She is best known for her landmark 2017 survey, "Efficient Processing of Deep Neural Networks," which has accumulated nearly 4,000 citations and has become an essential reference for researchers seeking to understand how to deploy computationally demanding AI models on resource-constrained hardware. Her work addresses a critical challenge in modern AI: while deep neural networks deliver remarkable accuracy across computer vision, speech recognition, and robotics, their computational cost makes real-world deployment difficult. Sze's research tackles this gap through co-designed hardware-software solutions, most notably the Navion chip — a remarkably power-efficient 2-mW accelerator enabling real-time visual-inertial odometry on nano drones and AR/VR devices. Beyond robotics navigation, she has contributed to embedded depth estimation through FastDepth and information-theoretic mapping algorithms that make autonomous exploration feasible on low-power platforms. Her recurring emphasis on bridging the gap between machine learning capability and practical hardware constraints makes her a foundational voice for engineers and researchers building the next generation of intelligent, edge-deployed systems.
Research Focus
Key Achievements
Top Papers
- 1Efficient Processing of Deep Neural Networks: A Tutorial and Survey3,979 citations · 2017
- 2
- 3Hardware for machine learning: Challenges and opportunities96 citations · 2018
- 4
- 5
- 6Efficient Processing of Deep Neural Networks: A Tutorial and Survey50 citations · 2017
- 7
- 8
- 9FastDepth: Fast Monocular Depth Estimation on Embedded Systems22 citations · 2019
- 10An Efficient and Continuous Approach to Information-Theoretic Exploration16 citations · 2020