Search Results - samuel+dillavou

2 Results Sort By:
A Robust Training Platform for Real-World Analog AI Hardware
Bias-tolerant training method for analog AI hardware that improves learning stability without digital correction. Problem: Analog computing could enable faster, lower-power AI systems, but real devices are difficult to train reliably because physical hardware is inherently imperfect. Small biases in sensing, signal application, or component behavior...
Published: 5/19/2026   |   Updated: 5/19/2026   |   Inventor(s): Andrea Liu, Samuel Dillavou, Douglas Durian, Marcelo Guzman
Keywords(s): Artificial Intelligence (AI) & Machine Learning, Autonomous Vehicles, Internet of Things
Category(s): Technology Classifications > Computer Information Systems, Technology Classifications > Hardware & Components, Technology Classifications > Robotics
Physical Neural Network for Fast, Scalable, and Power-Efficient Machine Learning Computation
A physical embodiment of a locally-learning neural network, which learns faster than current CPU-based nets. Problem: Machine learning is a ubiquitous tool that has found uses in engineering, science, and medicine. Neural networks are some of the most powerful machine learning algorithms, but they train very slowly when they become large. This hampers...
Published: 9/26/2025   |   Updated: 10/24/2024   |   Inventor(s): Samuel Dillavou, Douglas Durian, Andrea Liu, Menachem (Nachi) Stern, Marc Miskin
Keywords(s): Artificial Intelligence (AI) & Machine Learning, Data Science, Electronic Materials, Internet of Things, Nanotechnology, Software
Category(s): Technology Classifications > Computer Information Systems, Technology Classifications > Hardware & Components