Pattern-recognizing stochastic learning automata

By Andrew Barto, P. Anandan. 1985

How researchers see this paper

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

Citations in the corpus

Doubly Robust Policy Evaluation and Optimization (2014, background): “, 2002/03; Langford and Zhang, 2008), sometimes known as associative reinforcement learning (Barto and Anandan, 1985), are a natural generalization of the classic multiarmed ban-”
Counterfactual Estimation and Optimization of Click Metrics for Search Engines (2014, background): “The contextual-bandit formalism [2, 20] generalizes classic multi-armed bandits by introducing contextual information in the interaction loop between a learner and the environment it is situated in.”
Varieties of Helmholtz Machine (1996, methodology): “We have explored a reinforcement learning method (based on REINFORCE, Williams, 1992) which optimises both (cid:30) and (cid:18) during the wake phase (and therefore does without a sleep phase).”
Reinforcement Learning is Direct Adaptive Optimal Control (1992, background): “Extending this ideatoallow action selections todependonstate information introduces aspects offeedbackcontrol, pattern recognition, andassociative learning (e.g., [ 2 ,6]).”
Simple statistical gradient-following algorithms for connectionist reinforcement learning (1992, methodology): “It is interesting to compare this with the associative reward-penalty (An_e) algorithm (Barto, 1985; Barto & Anandan, 1985; Barto & Anderson, 1985; Barto & Jordan, 1987), which, for r E [0, 1], uses the learning rule where ~ is a positive learning rate parameter and 0 < X _< 1.”
Learning internal representations by error propagation (1986, background): “Another recent approach, also employing stochastic units, has been developed by Barto (1985) and various of his colleagues (cf. Barto & Anandan, 1985).”
Function Optimization using Connectionist Reinforcement Learning Algorithms (1991, methodology): “Furthermore, any reinforcement learning algorithm for adjusting the weights in the network (e.g., Barto, 1985; Barto & Anandan, 1985; Barto & Anderson, 1985; Munro, 1987; Sutton, 1984; Williams, 1986; 1987; 1988) can be used to provide a means of adapting the behavior of this trial point generator…”
Using Expectation-Maximization for Reinforcement Learning (1997, methodology): “Examples include the Keifer-Wolfowitz procedure (Wasan 1969), the ARP algorithm (Barto & Anandan 1985), and the REINFORCE framework (Williams 1992).”
Networks adjusting networks (1990, background): “Another candidate for the learning rule in step 4 is Barto and Anandan's AR P rule [2].”
Generative learning structures and processes for generalized connectionist networks (1993, background): “Alternatively, they may be obtained from not-so-specific feedback e.g., a reward/punishment signal used in reinforcement learning (Barto & Anandan, 1985), or they may be internally derived, e.g., based on an estimate of the output necessary from a node to…”
Reinforcement Learning: An Introduction (1998, methodology): “Barto and Anandan (1985) introduced an associative reinforcement learning algorithm called the associative reward-penalty (AR P) algorithm.”
The Helmholtz Machine (1995, background): “We define the energy of explanation to be E ( ; d) = log p( j )p(dj ; ) (2)”
Multiagent reinforcement learning in the Iterated Prisoner's Dilemma. (1996, background): “Speci cally, an agent's probability of selecting action ai from state s was p(ai) = eQ(s;ai)=t Pa2fC;Dg eQ(s;a)=t (5) where the temperature t was a function of the number n of PD games played so far: t = 5 0:999n: (6) If t < 0:01 then no exploration was performed, i.”
Connectionist expert systems (1988, methodology): “When selecting training examples (step (2) in the pocket algorithms, p.”