Papers from the 2011 AAAI Workshop
Joseph Modayil, Doina Precup, Satinder Singh, Workshop Cochairs
A truly intelligent agent should be capable of learning online from a lifetime of raw sensorimotor experience, by autonomously developing internal structures that provide the foundations for learning and further development. This problem sits at the core of artificial intelligence, but it has traditionally been difficult for algorithms to learn online from a single high-dimensional time series of correlated data. However, recent progress in several branches of AI suggests that the goal is getting closer. This workshop will bring together researchers to share ideas and insights from several subfields including reinforcement learning, robotics, and deep learning. The workshop will cover methodology for life-long learning and discuss experimental platforms that can be used towards this goal.