all InfoSec news
Exploration is Harder than Prediction: Cryptographically Separating Reinforcement Learning from Supervised Learning
April 8, 2024, 4:11 a.m. | Noah Golowich, Ankur Moitra, Dhruv Rohatgi
cs.CR updates on arXiv.org arxiv.org
Abstract: Supervised learning is often computationally easy in practice. But to what extent does this mean that other modes of learning, such as reinforcement learning (RL), ought to be computationally easy by extension? In this work we show the first cryptographic separation between RL and supervised learning, by exhibiting a class of block MDPs and associated decoding functions where reward-free exploration is provably computationally harder than the associated regression problem. We also show that there is …
arxiv cryptographic cs.cc cs.cr cs.ds cs.lg easy exploration extension practice prediction work
More from arxiv.org / cs.CR updates on arXiv.org
Jobs in InfoSec / Cybersecurity
Offensive Security Engineering Technical Lead, Device Security
@ Google | Amsterdam, Netherlands
Senior Security Engineering Program Manager
@ Microsoft | Redmond, Washington, United States
Information System Security Analyst
@ Resource Management Concepts, Inc. | Dahlgren, Virginia, United States
Critical Facility Security Officer - Evening Shift
@ Allied Universal | Charlotte, NC, United States
Information System Security Officer, Junior
@ Resource Management Concepts, Inc. | Patuxent River, Maryland, United States
Security Engineer
@ JPMorgan Chase & Co. | Plano, TX, United States