RESEARCH
Analog in-memory computing isn't new—it's been developed in research labs for decades. Here's the science we're building on and the people doing the work.
We believe in transparency about where we're starting from. This page documents the foundational research, key papers, and leading groups in analog computing. These are the shoulders we're standing on—and in many cases, the people we're talking to about joining or advising.
FOUNDATIONAL PAPERS
Sebastian, A., Le Gallo, M., Khaddam-Aljameh, R. & Eleftheriou, E.
Nature Nanotechnology, 2020
The definitive review from IBM Research Zurich. Covers memristive devices, phase-change memory, and the computational primitives they enable.
Gokmen, T. & Vlasov, Y.
Nature, 2016
IBM's proof that analog crossbar arrays can train neural networks without accuracy loss. Key milestone for the field.
Yao, P., Wu, H., Gao, B., et al.
Nature, 2020
Tsinghua/Stanford collaboration demonstrating a full CNN on memristor hardware. Achieved image recognition with real analog devices.
Le Gallo, M., et al.
Nature Electronics, 2023
IBM's most recent chip: 64 analog compute cores achieving record efficiency. The closest thing to a commercial-ready analog AI processor.
EVENTS & CHALLENGES
HARDWARE & DEVICES
INIVATION (DVS, DAVIS)
Pioneer sensors from Institute of Neuroinformatics. DVS128, DAVIS240/346 series.
PROPHESEE METAVISION
High-resolution event cameras with advanced software ecosystem. Up to 1280x720 resolution.
RESEARCH APPLICATIONS
SLAM & VISUAL ODOMETRY
Event-based simultaneous localization and mapping using temporal contrast changes for robust navigation.
METHODS: direct methods, feature tracking, visual-inertial fusion
OPTICAL FLOW & MOTION
High-speed motion field estimation using event streams for real-time perception.
USES: drone navigation, collision avoidance, motion segmentation
3D RECONSTRUCTION
Monocular and stereo depth estimation using event cameras with structured light fusion.
TECHNIQUES: contrast maximization, deep learning, semi-dense
OBJECT RECOGNITION
Pattern recognition and object tracking using sparse event representations.
METHODS: HOTS, graph neural networks, attention mechanisms
SPACE
Satellite tracking, debris monitoring, lunar nav
AUTOMOTIVE
Driver monitoring, lane detection, collision
BIOMEDICAL
Retinal implants, eye tracking, neural interfaces
INDUSTRIAL
Quality control, vibration analysis, monitoring
KEY DATASETS
SOFTWARE & TOOLS
LEADING RESEARCH GROUPS
INI
UZH & ETH Zurich
RPG
University of Zurich
EDPR
IIT Italy
GRASP LAB
UPenn
INTEL LABS
Loihi Development
UMD
Perception & Robotics
ICNS
Western Sydney
PEKING UNIVERSITY
Camera Intelligence Lab
KAIST
Visual Intelligence Lab
EDUCATIONAL RESOURCES
RESOURCES
This directory draws from the community-driven Event-based Vision Resources repository.