Introduction to Reinforcement Learning

By Richard S. Sutton, Andrew Barto. 1998

How researchers see this paper

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

In Reinforcement Learning

Citations in the corpus

The Arcade Learning Environment: An Evaluation Platform for General Agents (2012, methodology): “A detailed explanation of SARSA(λ) and its extensions can be found in the work of Sutton and Barto (1998).”
Asynchronous Methods for Deep Reinforcement Learning (2016, methodology): “The target value used by one-step Sarsa is r + γQ ( s (cid:48) , a (cid:48) ; θ − ) where a (cid:48) is the action taken in state s (cid:48) (Rummery & Niranjan, 1994; Sutton & Barto, 1998).”
Learning to Extract Coherent Summary via Deep Reinforcement Learning (2018, methodology): “E y~ t;~s t˘ˇ h tR~ trlogˇ (~y tj~s t) i ; (2) where ~s tand y~ tare randomly sampled from ˇ , R~ tis the actual return received since ~s t and y~ t. A detailed proof of Equation 2 could be found in (Sutton and Barto 1998) and is omitted for brevity here. The parameters are updated as follows: + tR~”
Reinforcement Learning with Deep Energy-Based Policies (2017, background): “If the objective function is the expected discounted sum of rewards, the policy improvement theorem (Sutton & Barto, 1998) describes how policies can be improved monotonically.”
The Predictron: End-To-End Learning and Planning (2016, methodology): “This λ-preturn is analogous to the λ-return in the forward-view TD(λ) algorithm (Sutton, 1988; Sutton & Barto, 1998).”
Decoupled Neural Interfaces using Synthetic Gradients (2016, methodology): “Mixing real and synthetic gradients results inBP (λ), an algorithm anolgous to TD(λ) for reinforcement learning (Sutton & Barto, 1998).”
Deep learning in neural networks: An overview (2014, background): “This assumption does not hold in the broader fields of Sequential Decision Making and Reinforcement Learning (RL) (Hutter, 2005; Kaelbling, Littman, & Moore, 1996; Sutton & Barto, 1998; Wiering & van Otterlo, 2012) (Section 6).”
Learning World Graphs to Accelerate Hierarchical Reinforcement Learning (2019, positive): “at has a feasible connection to g win G w. 2. Planning: We estimate the optimal traversal route from g0 wto g based on the G edge weights. Here we use the classic dynamic programming planning methods [85, 27], although other (learned) methods can be applied. 3. Execution: once the route is planned, ”
Toward Interpretable Deep Reinforcement Learning with Linear Model U-Trees (2018, methodology): “e s, as shown in the Leaf Node 5 of Figure 3. So LMUT builds a Markov Decision Process (MDP) from the interaction data between environment and deep model. Compared to a linear Q-function approximator [18], a LMUT denes an ensemble of linear Q-function models, one for each partition cell. Since each ”
Reinforcement learning in robotics: A survey (2013, background): “One example is tile coding (Sutton and Barto, 1998), where the space is subdivided into (potentially irregularly shaped) regions, called tiling.”
An Emphatic Approach to the Problem of Off-policy Temporal-Difference Learning (2015, background): “(16) The λ-return plays a key role in the theoretical understanding of TD methods, in particular, in their forward views (Sutton & Barto 1998, Sutton, Mahmood, Precup & van Hasselt 2014).”