An intrinsic reward mechanism for efficient exploration

By Özgür Simsek, Andrew Barto. 2006

136 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Reinforcement Learning

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

In Computer Networks and Communications

In Reinforcement Learning

Citations in the corpus

Learning Parameterized Skills (2012, methodology): “This is also the approach taken by Şimşek and Barto [15].”
Unifying Count-Based Exploration and Intrinsic Motivation (2016, background): “, 2015), value error (Simsek and Barto, 2006), learning progress (Schmidhuber, 1991), or mutual information (Still and Precup, 2012; Mohamed and Rezende, 2015).”
Exploration from Demonstration for Interactive Reinforcement Learning (2016, background): “Using an internal reward function [9] is an interesting approach to exploration in RL.”
Surprise and Curiosity for Big Data Robotics (2014, background): “Typically intrinsically motived systems are segregated into two camps: 1) knowledge-driven sequential task learning (Schmidhuber 1991, Oudeyer et al 2007), and 2) competence-driven option or subtask training (Singh et al 2005, Simsek and Barto 2006, Schembri et al 2007).”
Emerging social awareness: Exploring intrinsic motivation in multiagent learning (2011, background): “include the hierarchical acquisition of skills using intrinsically motivated reinforcement learning [5]-[7], the acquisition of a visual-attention system from motivation variables [8], and others [3], [4], [9], [10].”
Intrinsically motivated information foraging (2010, background): “In [17], an intrinsic exploration policy selects actions that are expected to maximize positive learning updates in a task-specific value function.”
Transfer Learning for Reinforcement Learning Domains: A Survey (2009, background): “While no current TL algorithms take such an approach, there has been some work on the question of learning a policy that is exploitable (without attempt to maximize the on-line reward accrued while learning) in non-transfer contexts (Şimşek and Barto, 2006).”
Algorithms for Reinforcement Learning (2010, background): “Some experimental results (for some heuristic algorithms) can be found in the paper by Şimşek and Barto (2006).”
Exploration from Generalization Mediated by Multiple Controllers (2013, methodology): “The rather more ambitious and distinctly less complete goal is to support attempts to bring intrinsic motivational concepts such as curiosity and uncertainty back under the fold of conventional control in terms of sophisticated exploration bonuses (Şimşek and Barto 2006).”
Full Paper Reinforcement Learning for Imitating Constrained Reaching Movements (unknown): “…which is based on function approximation and gradient descent algorithms [9–13] More recently, studies on the eﬀect of the reward function on the speed of learning [27], on the eﬀect of using inaccurate model [28] or on new eﬃcient exploration process [29] have been conducted in the RL ﬁeld.”
Real time targeted exploration in large domains (2010, background): “Simsek and Barto [10] take a similar approach, examining”
Competence progress intrinsic motivation (2010, methodology): “[16] cite [10] but do not mention this connection.”