Reinforcement learning
By Florentin Wörgötter, Bernd Porr. 2019
3,375 citations recorded by the source corpus.
The discussion here considers a much more common learning condition where an agent, such as a human or a robot, has to learn to make decisions in the environment from simple feedback. Such feedback is provided only after periods of actions in the form of reward or punishment without detailing which of the actions has contributed to the outcome. This type of learning scenario is called reinforcement learning. This learning problem is formalized in a Markov decision-making process with a variety of related algorithms. The second part of this chapter will use function approximators with neural networks which have made recent progress as deep reinforcement learning.
Subfields: Artificial Intelligence, Developmental and Educational Psychology, 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
- Sergey Levine
- Yoshua Bengio
- Pieter Abbeel
- Jianfeng Gao
- Mohit Bansal
- Ruslan Salakhutdinov
- Percy Liang
- Quoc V. Le
In Developmental and Educational Psychology
- J. Urgen Schmidhuber
- Laura Schulz
- Noah D. Goodman
- Samuel J. Gershman
- Petar Milin
- Michael Ramscar
- John P. O'Doherty
- Raphael Fonteneau
In Reinforcement Learning
Citations in the corpus
- Recurrent World Models Facilitate Policy Evolution (2018, background): “For many reinforcement learning (RL) problems [37, 96, 106], an artificial RL agent may also benefit from a predictive model (M) of the future [95, 104] (model-based RL).”
- Value-Decomposition Networks For Cooperative Multi-Agent Learning (2017, methodology): “We recall some key concepts of the RL setting (Sutton and Barto, 1998), an agent-environment framework (Russell and Norvig, 2010) in which an agent sequentially interacts with the environment over a sequence of timesteps, t = 1 , 2 , 3 , . . . , by executing actions and receiving observations and…”
- Deep learning in neural networks: An overview (2014, neutral): “Compare RBM-based RL (Sallans & Hinton, 2004) with high-dimensional inputs (Elfwing, Otsuka, Uchibe, & Doya, 2010), earlier RL Atari players (Grüttner, Sehnke, Schaul, & Schmidhuber, 2010), and an earlier, raw video-based RL NN for computer games (Koutník, Cuccu, Schmidhuber, & Gomez, 2013) trained…”
- Solving Rubik's Cube with a Robot Hand (2019, neutral): “Related ideas were also discussed in [27, 11, 117, 20].”
- World Models (2018, background): “In many reinforcement learning (RL) problems (Kaelbling et al., 1996; Sutton & Barto, 1998; Wiering & van Otterlo, 2012), an artificial agent also benefits from having a good representation of past and present states, and a good predictive model of the future (Werbos, 1987; Silver, 2017), preferably…”
- Reinforcement Learning in Healthcare: A Survey (2019, background): “A more comprehensive and in-depth review on these issues can be found in Reference [238], and more recently in References [126, 213].”
- Tag-Trigger-Consolidation: A Model of Early and Late Long-Term-Potentiation and Depression (2008, neutral): “[52] which can be interpreted as a synapse specific tag.”
- Policy Networks with Two-Stage Training for Dialogue Systems (2016, background): “A good selection of Ba, which is called the baseline, can reduce the variance dramatically (Sutton and Barto, 1998).”
- Hybrid Reinforcement/Supervised Learning of Dialogue Policies from Fixed Data Sets (2008, methodology): “There have been some proposals in RL for learning a policy that is different from that used to generate the data (called “off-policy” learning), but these methods have been found not to work well with linear function approximation (Sutton and Barto 1998).”
- Batch Mode Reinforcement Learning based on the Synthesis of Artificial Trajectories (2012, positive): “Over the last decade, techniques developed by the Re-inforcement Learning (RL) community (Sutton and Barto 1998) have become more and more popular for addressing those types of problems.”
- Natural Language Generation as Planning Under Uncertainty for Spoken Dialogue Systems (2009, methodology): “Reinforcement Learning (RL) provides a principled, data-driven optimisation framework for our type of planning problem (Sutton and Barto, 1998).”
- One Big Net For Everything (2018, background): “Unlike in the present paper, predictions of expected cumulative rewards are actually essential in traditional reinforcement learning [29, 90, 101, 76] where they are usually limited to the case of scalar rewards (while ONE’s rewards may be vector-valued like in old work of 1990 [55, 56]).”
- Using Decision Lists to Construct Interpretable and Parsimonious Treatment Regimes (2015, positive): “For nonparametric Q-learning, we use support vector regression when Y is continuous and support vector machines when Y is binary (Zhao et al., 2011), both are implemented using a Gaussian kernel.”
- WRITING MEMORIES WITH LIGHT-ADDRESSABLE REINFORCEMENT CIRCUITRY (2009, neutral): “…normalized to the number of electric shocks consumed. p < 0.0001; Kruskal-Wallis ANOVA; **, significantly different from 2 min of Pavlovian conditioning in post hoc comparison (n = 20 flies per condition; means ± SEM). possible genetically to separate reinforcement from locomotor circuits.”
- Bandit Structured Prediction for Neural Sequence-to-Sequence Learning (2017, methodology): “(2016) apply techniques from reinforcement learning (Sutton and Barto, 1998; Sutton et al., 2000) and imitation learning (Schaal, 1999; Ross et al.”
- Toward an Integration of Deep Learning and Neuroscience (2016, methodology): “The “attention-gated reinforcement learning” (AGREL) networks of Stanisor et al. (2013), Brosch et al. (2015), and Roelfsema and van Ooyen (2005), and variants like KickBack (Balduzzi, 2014), give a way to compute an approximation to the full gradient in a reinforcement learning context using a…”