Neural Networks for Control

By Eduardo Sontag. 1993

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

An Actor-Critic Algorithm for Sequence Prediction (2016, methodology): “While RL methods with non-linear function approximators are not new (Tesauro, 1994; Miller et al., 1995), they have recently surged in popularity, giving rise to the field of ‘deep RL’ (Mnih et al.”
Deep learning in neural networks: An overview (2014, methodology): “1 RL Through NN World Models Yields RNNs With Deep CAPs In the special case of an RL NN controller C interacting with a deterministic environment, a separate FNN called M can learn to become C’s world model through system identification, predicting C’s inputs from previous actions and inputs (e.g., ”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, methodology): “The first combination of an RL RNN C and an UL RNN M was ours and dates back to 1990 [223, 222, 226, 227], generalizing earlier similar controller/model systems (CM systems) based on FNNs [298, 179]; compare related work [177, 119, 301, 300, 209, 120, 178, 302, 73, 45, 144, 166, 153, 196, 60][245, S”
Deep learning in robotics: a review of recent research (2017, background): “that was needed for operating dynamical systems in continuous spaces [13,14], and closely related fuzzy systems seemed well suited for nominal logical control decisions [15].”
On the computational power of circuits of spiking neurons (2004, background): “the only mathematical results that provide such bounds (see [31] for a discussion) are almost never used practically.”
Structure-Adaptable Digital Neural Networks (1999, methodology): “Arti cial neural networks have been proposed as an alternative for implementing adaptive controllers which are capable of learning [138, 226].”
Sobolev Training for Neural Networks (2017, methodology): “Similar techniques have also been used in critic based Reinforcement Learning (RL), where a critic’s derivatives are trained to match its target’s derivatives [29, 15, 5, 4, 26] using small, sigmoid based models.”
Self-Organizing Feature Maps for Modeling and Control of Robotic Manipulators (2003, background): “These properties make neural networks very attractive for adaptive robot control (Miller et al., 1990; Hunt et al., 1992; Balakrishnan and Weil, 1996), by learning the mapping between various movement variables without an accurate knowledge of the system parameters or the analytical equations govern”
The Artificial Life Roots of Artificial Intelligence (1993, background): “Reinforcement learning methods have been shown to be capable of impressive learning behavior in simulations or engineering contexts [81], but there are again serious difficulties in the application to physical autonomous agents.”
Sequence Learning (2001, background): “Concerning computational power, among positive results we can mention that recurrent networks can model any first order discrete-time, time invariant, non-linear system (see for example [54] [59]).”
Understanding and Modeling the Human Driver (2003, background): “Neural network models [116–126] of the driving process essentially subsume many of these ideas within the neural net architecture and associated connection weights.”
Hoeffding Races: Accelerating Model Selection Search for Classification and Function Approximation (1993, background): “This paper focuses on the special case of leave-one-out cross validation applied to memory-based learning algorithms, but we also argue that it is applicable to any class of model selection problems.”
Combining the benefits of function approximation and trajectory optimization (2014, background): “Indeed control was among the early applications of neural networks [17], and there is still intense interest in this topic in multiple communities.”