The many faces of optimism: a unifying approach

By Istvan Szita, András Lörincz. 2008

105 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Lihong Li
2. Sergey Levine
3. Hado van Hasselt
4. David Silver
5. Pieter Abbeel
6. Shimon Whiteson
7. Michael L. Littman
8. R´emi Munos

In Computational Theory and Mathematics

1. Csaba Szepesvari
2. Lihong Li
3. Michael L. Littman
4. Hado van Hasselt
5. Romain Laroche
6. David Silver
7. Carlos Diuk
8. Thomas J. Walsh

Citations in the corpus

• Safe Policy Improvement with Baseline Bootstrapping (2017, background): “Reinforcement Learning (RL, [28]) consists in discovering through optimism in the face of uncertainty (OFU [30]), in an unknown uncertain environment, which action is the most valuable in a particular situation.”
• Model-based reinforcement learning with nearly tight exploration complexity bounds (2010, methodology): “The first such algorithm was E 3 (Kearns & Singh, 2002), which was followed by other, simpler algorithms like Rmax (Brafman & Tennenholtz, 2002), model-based intervalestimation (MBIE) (Strehl & Littman, 2005) and optimistic initial model (OIM) (Szita & L˝ orincz, 2008), all of which are model-based.”
• Sample Complexity Bounds of Exploration (2012, background): “Furthermore, it is possible to replace the constant optimistic value Vmax by a non-constant, optimistic value function to gain further improvement in the sample complexity bound (Strehl et al, 2009; Szita and Lőrincz, 2008).”
• Intrinsically Motivated Learning of Real-World Sensorimotor Skills with Developmental Constraints (2013, unknown): “, 1994; Thrun, 1992, reinforcement learning (e.g. Barto et al., 2004; Brafman and Tennenholtz, 2001; Schmidhuber, 1991; Sutton, 1990; Szita and Lorincz, 2008), computational neuroscience (e.”
• Exploring compact reinforcement-learning representations with linear regression (2009, background): “This property follows from standard arguments [17]: for unknown states, the noise term of the prediction error can be bounded by Azuma’s inequality, and R0 can be set high enough so that the second term is positive and dominates.”
• Deep Reinforcement Learning with Double Q-Learning (2015, background): “The overestimations should not be confused with optimism in the face of uncertainty (Sutton, 1990; Agrawal, 1995; Kaelbling et al., 1996; Auer et al., 2002; Brafman and Tennenholtz, 2003; Szita and Lőrincz, 2008; Strehl et al., 2009), where an exploration bonus is given to states or actions with unc”