Supervised learning and systems with excess degrees of freedom

By Michael I. Jordan. 1988

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

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”
Neural network control for a closed-loop System using Feedback-error-learning (1993, methodology): “Thus, Jordan proposed forward-inverse-modeling (Jordan, 1988) and Albus (1975), Miller (Miller, Hewes, Glanz, & Kraft, 1990), Atkeson (Atkeson & Reinkensmeyer, 1988), Psaltis (Psaltis et al.”
NeuroAnimator: fast neural network emulation and control of physics-based models (1998, background): “More recently, Jordan and Rumelhart [9] proposed a two step approach to learning controllers for physical robots.”
Learning Algorithms for Networks with Internal and External Feedback (1990, methodology): “In diierent contexts and with diierent degrees of generality, this basic principle for credit assignment based on system identiication has been previously described in (Werbos, 1977), (Jordan, 1988), (Munro, 1987), (Robinson and Fallside, 1989), (Nguyen and Widrow, 1989), and (Schmidhuber, 1990b).”
Learning to Control an Unstable System with Forward Modeling (1989, methodology): “Previous wor k within the forward modeling paradigm focused on models of fixed kinematic or dynamic properties of the controlled plant (Jordan, 1988,1990; Jordan &, Rumelhart, 1990).”
Making the World Di erentiable: On Using Self-Supervised Fully Recurrent Neural Networks for Dynamic Reinforcement Learning (1990, background): “Werbos, Jordan, Widrow In 8], Jordan uses a model network for constructing a mapping from output actions of a control network to the environmental eeects caused by these actions.”
Forward Models: Supervised Learning with a Distal Teacher (1992, methodology): “Although the backpropagation algorithm never forms this matrix explicitly (backpropagation is essentially a factorization of the matrix; Jordan, 1988), Equation 6 nonetheless describes the results of the computation performed by backpropagation.”
A possibility for implementing curiosity and boredom in model-building neural controllers (1991, background): “6 References[1] M. I. Jordan.”
Making the world differentiable: on using self supervised fully recurrent neural networks for dynamic reinforcement learning and planning in non-stationary environments (1990, background): “In the context of environments where the environmental state is fully determined by the current input, Jordan [8] al!io notes that a model network does not need to be perfect to allow increasing performance of the control”
Reinforcement Learning Upside Down: Don't Predict Rewards - Just Map Them to Actions (2019, background): “Using SL for certain aspects of RL dates back to the 1980s and 90s [70, 31, 22, 73, 72, 39, 32].”
Learning to Generate Artificial Fovea Trajectories for Target Detection (1991, background): “The reason is: It su ces if the inner products of the approximated gradients (based on an inaccurate model) for C and the true gradients (according to a hypothetical perfect model) are always positive (see also [2]).”
A Self-Organizing Neural Model of Motor Equivalent Reaching and Tool Use by a Multijoint Arm (1993, methodology): “For exarnple, Jordan (1988) described a model incorporating a temporal smoothness constraint which helps to ensure that temporally adjacent actions do not conflict with each other.”