Communications of the ACM
To Learn to Deal With Uncertainty, This AI Plays Pong
Researchers at the Massachusetts Institute of Technology developed a new algorithm that enables artificial intelligence models to account for uncertain data.
Credit: iStockphoto
Massachusetts Institute of Technology (MIT) researchers have developed a new algorithm that enables artificial intelligence (AI) models to account for uncertain data.
The researchers developed a secondary algorithm, Certified Adversarial Robustness for deep RL (CARRL), that can be built on top of an existing reinforcement learning (RL) model.
Said MIT's Michael Everett, "Our key innovation is that rather than blindly trusting the measurements, as is done today [by AI models], our algorithm CARRL thinks through all possible measurements that could have been made, and makes a decision that considers the worst-case outcome."
The researchers tested CARRL on such tasks as collision avoidance simulations and the classic computer game Atari pong, and found that CARRL improved the RL model's ability to compensate for potential inaccurate or "noisy" data.
From IEEE Spectrum
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found