A stochastic reinforcement learning algorithm for learning real-valued functions

By Vijaykumar Gullapalli. 1990

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

End-to-End Training of Deep Visuomotor Policies (2015, neutral): “d in two conference papers (Levine and Abbeel, 2014; Levine et al., 2015), which we extend to introduce visual input into the policy. 2. Related Work Reinforcement learning and policy search methods (Gullapalli, 1990; Williams, 1992) have been applied in robotics for playing games such as table tenn”
Reinforcement Learning: A Survey (1996, neutral): “Gullapalli (1990, 1992) has developed a \neural" reinforcement-learning unit for use in continuous action spaces.”
Online Model Learning Algorithms for Actor-Critic Control (2015, methodology): “2 Actor-Only Methods and the Policy Gradient Policy gradient methods, such as the stochastic real-valued (SRV) algorithm (Gullapalli, 1990) and REINFORCE (Williams, 1992) algorithms, are principally actor-only and do not use any form of a stored value function.”
Reinforcement Learning by Value Gradients (2008, methodology): “The stochastic realvalued (SRV) unit algorithm (Gullapalli, 1990) is used to train the actor while efficiently compensating for the added noise: ∆~z = αnt ( ∂π ∂~z )”
Reinforcement Learning in Continuous Time and Space (2000, background): “The size of the perturbation σ was tapered off as the performance improved (Gullapalli, 1990).”
Biped dynamic walking using reinforcement learning (1997, methodology): “I am forever grateful and will never forget to share what they have so generously given me. vi TABLE OF CONTENTS INTRODUCTION 12 BIPED WALKING 15 1 HISTORY 15 2 STATIC WALKING 16 3 DYNAMIC WALKING 17 4 INVERTED PENDULUM MODEL 19 5 CPG METHODS 20 6 QUICK OVERVIEW OF EXISTING BIPED ROBOTS 21 7 SUMMARY”
Neural mechanism for stochastic behaviour during a competitive game (2006, methodology): “In order to check this hypothesis we implemented a modified version of the meta-learning algorithm proposed by Schweighofer and Doya (2003), which is an extension of the stochastic real value units algorithm (Gullapalli, 1990).”
Recurrent policy gradients (2010, background): “…the obvious combination with standard backpropagation techniques has not been extensively investigated (a notable exception being the SRV algorithm [15, 16], which was, however, solely applied to feedforward networks In this paper, we address this shortcoming, and show complex neural networks with…”
Robust Reinforcement Learning (2005, unknown): “The sizes of the perturbation σu and σw were tapered off as the performance improved ( Gullapalli, 1990 ).”
Simple statistical gradient-following algorithms for connectionist reinforcement learning (1992, methodology): “It is interesting to compare REINFORCE for such a unit with an alternative algorithm for the adaptation of # and a that has been proposed by Gullapalli (1990).”
Ieee Transactions on Systems, Man and Cybernetics—part C: Applications and Reviews 1 a Survey of Actor-critic Reinforcement Learning: Standard and Natural Policy Gradients (unknown): “In [52], it is noted that, in light of the policy gradient theorem that was only published many years later, Gullapalli's earlier idea in [32] of using the temporal difference δ in the gradient used to update the policy weights can be shown to yield the true policy gradient ∇ ϑ J(ϑ), and hence…”
On the Control of Control: The Role of Dopamine in Regulating Prefrontal Function and Working Memory (2007, background): “…consistent with the practice in other reinforcement learning simulations of having noise levels inversely related to the level of reward predicted (Gullapalli 1990).9 Results Ten runs of the simulation were performed, each with randomly assigned initial weights for the modifiable connections in…”
Reinforcement learning of motor skills with policy gradients (2008, methodology): “Without a formal derivation of it, the policy gradient theorem has been applied in Gullapalli (1990, 1992) and Kimura and Kobayashi (1997) using estimated value functionsQ (x,u) instead of the term ∑H l=k alrl and a baseline bk = V (xk, k). Note that the version introduced in Kimura and Kobayashi (1”
A comparison of direct and model-based reinforcement learning (1997, background): “A number of researchers have focused on this case (Gullapalli, 1990; Jordan and Jacobs, 1990; Williams, 1990; Millington, 1991; Baird and Klopf, 1993; Bradtke, 1993; Bradtke et al., 1994; Doya, 1996).”