Explanation-Based Neural Network Learning for Robot Control

By Tom Michael Mitchell, Sebastian Thrun. 1992

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

Meta Networks (2017, background): “By following the success of the previous work (Mitchell et al., 1993; Younger et al., 1999; Andrychowicz et al., 2016; Ravi & Larochell, 2017), we studied the meta information in the form of the loss gradient of neural nets.”
Conservative Q-Improvement: Reinforcement Learning for an Interpretable Decision-Tree Policy (2019, methodology): “We evaluate our approach on the RobotNav environment from [Mitchell and Thrun, 1993].”
Learning Rapid-Temporal Adaptations (2017, background): “There is growing interest and progress in building ﬂexible, adaptive models, particularly within the framework of metalearning , or learning to learn [24, 1, 42, 2].”
Knowledge Matters: Importance of Prior Information for Optimization (2013, methodology): “An EBNN uses previously learned domain knowledge as an initialization or search bias (i.e. to constrain the learner in the parameter space) (O’Sullivan, 1996; Mitchell and Thrun, 1993).”
A lifelong learning perspective for mobile robot control (1994, methodology): “In this figure, we visualized several learning episodes viewed from Sebastian Thrun A Lifelong Learning Perspective for Mobile Robot Control 6 a bird eye’s view, using a sonar map techniques described in [27].”
Integrated perception with recurrent multi-task neural networks (2016, methodology): “While MLT trains many tasks in parallel, Mitchell and Thrun [18] propose a sequential transfer method called Explanation-Based Neural Nets (EBNN) which exploits previously learnt domain knowledge to initialise or constraint the parameters of the current task.”
Towards Knowledgeable Supervised Lifelong Learning Systems (2020, background): “Acknowledged for the first time more than two decades ago (Mitchell & Thrun, 1993; Thrun, 1996), lifelong learning has been defined as a continuous learning process that relies on the execution of a sequence of related learning tasks (Chen & Liu, 2016).”
Robust Learning with Jacobian Regularization (2019, background): “8 Different incarnations of a similar idea have appeared in the following decades [Simard et al., 1992, Mitchell and Thrun, 1993, Aires et al., 1999, Rifai et al., 2011, Gulrajani et al., 2017, Yoshida and Miyato, 2017, Czarnecki et al., 2017, Jakubovitz and Giryes, 2018].”
Issues in Using Function Approximation for Reinforcement Learning (1999, methodology): “the Backpropagation algorithm [4, 8, 9 ]. In some cases learning failed since the function approximator at hand was not capable of representing reasonable value functions at a ll [13].”
Is Learning The n-th Thing Any Easier Than Learning The First? (1995, methodology): “The last method described here uses the explanation-based neural network learning algorithm (EBNN), which was originally proposed in the context of reinforcement learning [8, 17].”
Combining Symbolic and Neural Learning (1994, background): “(Recently, Mitchell and Thrun (1993) proposed an explanation-based, though non-symbolic, method for training neural networks for reinforcement learning tasks.”
Knowledge-Based Artificial Neural Networks (1994, background): “The focus of these systems has varied from reducing the number of examples required for robot learning [29], to probabilistic logics [24], to the propagation of Mycinlike certainty factors [21], to the refinement of systems for fuzzy logic [5].”
Rapid Adaptation with Conditionally Shifted Neurons (2017, background): “Other previous work on metalearning has also formulated the problem as two-level learning: specifically, “slow” learning of a meta model across several tasks, and “fast” learning of a base model that acts within each task (Schmidhuber, 1987; Bengio et al., 1990; Hochreiter et al., 2001; Mitchell et ”