Fast Online Policy Gradient Learning with SMD Gain Vector Adaptation

By N. Schraudolph, Douglas Aberdeen, Jin Yu. 2005

42 citations recorded by the source corpus.

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

Painless Stochastic Gradient: Interpolation, Line-Search, and Convergence Rates (2019, methodology): “These techniques include using meta-learning procedures to modify the main stochastic algorithm [6, 90, 77, 2, 65, 90, 79], heuristics to adjust the learning rate on the ﬂy [44, 21, 72, 76], and recent adaptive methods inspired by online learning [22, 91, 40, 69, 62, 70, 53].”
Online Learning Rate Adaptation with Hypergradient Descent (2017, methodology): “ate of a main algorithm by using an averaging algorithm based on the mean of a sequence of adapted learning rates, also investigating rates of convergence. The stochastic meta-descent (SMD) algorithm [23, 22], developed as an extension of the gain adaptation work by Sutton [27], operates by multipli”
Solving Deep Memory POMDPs with Recurrent Policy Gradients (2007, unknown): “After being introduced by Williams [6] and Gullapalli [7], they have recently been extended to deal more efficiently [8] with complex high-dimensional tasks [9].”
Parameter-exploring policy gradients (2010, unknown): “…the policy directly from a value function, are among the few feasible optimisation strategies for complex, high dimensional reinforcement learning problems with continuous states and actions Benbrahim and Franklin (1997); Peters and Schaal (2006); Schraudolph et al. (2006); Peters et al. (2005).”
Policy Gradients with Parameter-Based Exploration for Control (2008, background): “Policy gradient methods, so called because they search in policy space without using value estimation, are among the most effective optimisation strategies for complex, high dimensional reinforcement learning tasks [1–4].”