New AI Theory Could Make Your Phone Smarter and Cheaper
This could mean faster, more private AI on your devices without the cloud.
Steinke and Zakynthinou(2020) introduces the Conditional Mutual Information (CMI) framework of analyzing the information complexity of learning algorithms based on algorithm-dependent information-theoretic quantities. We study one of these quantities, the evaluated Conditional Mutual Information (eCMI).
It has been an interesting question whether the optimal PAC guarantee for VC classes can be recovered from the algorithm-dependent analyses via CMI. And we show that it is possible to recover this guarantee by constructing a learning algorithm whose eCMI is of order O(d) in the realizable case, where d is the VC-dimension of the concept class. Specially, our algorithm is a randomized Majority-of-5 base learners with optimal in-expectation generalization guarantee.
- Researchers found the most efficient way for AI to learn from examples, using a measure called information complexity.
- This could lead to AI that runs on your device, saving data, battery, and improving privacy.
- The result is theoretical for now, so real-world apps may take years to appear.
Why It Matters
More efficient AI means faster, private, offline smart features on your phone and gadgets.