Planning simple trajectories using neural subgoal generators

By J. Urgen Schmidhuber, Reiner Wahnsiedler. 1993

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

Deep learning in neural networks: An overview (2014, background): “In particular, gradient-based subgoal discovery with FNNs or RNNs decomposes RL tasks into subtasks for RL submodules (Schmidhuber, 1991b; Schmidhuber & Wahnsiedler, 1992).”
One Big Net For Everything (2018, background): “Also in 1990, gradient-based recurrent subgoal generators [57, 58, 80] used special start and goal-deﬁning input vectors, also for an evaluator network predicting the costs and rewards associated with moving from starts to goals.”
CHILD: A First Step Towards Continual Learning (1997, methodology): “The importance of hierarchy in adaptive systems that perform temporal tasks has been noted often, and many systems have been proposed in which hierarchical architectures are developed by hand top down as an efficient method for modularizing large temporal tasks (Albus, 1979; Dayan and Hinton, 1993; ”
Learning to Learn (1998, unknown): “An important collection of papers on the topic of metaphor can be found in Ortony (1993). Other linguistic mechanisms which exploit the transfer of "meaning" or "semantic content" from one situation or utterance to another include comparison, analogy (defined here in a constrained linguistic context”
Why Does Hierarchy (Sometimes) Work So Well in Reinforcement Learning? (2019, background): “…an agent learns to solve tasks from trial-and-error experience, and the use of hierarchical reinforcement learning (HRL) has long held the promise to elevate the capabilities of RL agents to more complex tasks (Dayan & Hinton, 1993; Schmidhuber, 1993; Parr & Russell, 1998; Barto & Mahadevan, 2003).”
Continual learning in reinforcement environments (1995, methodology): “In existing systems such as those of Albus [1], Roitblat [86, 87], Jameson [42], Lin [55], Wixson [126], Dayan and Hinton [21], Schmidhuber and Wahnsiedler [97], and Singh [99], hierarchical architectures are developed top down by hand as an e cient method for modularizing large temporal tasks.”
Formal Theory of Creativity, Fun, and Intrinsic Motivation (1990–2010) (2010, methodology): “apparently being unaware of previous work on hierarchical R L systems that can discover their own subgoals [60], [102], [6 1], [118], [79], [1].”
Maximizing Fun by Creating Data with Easily Reducible Subjective Complexity (2013, background): “Some goal-seeking RL systems search a space of possible subgoal combinations, internally rewarding subsystems whose policies learn to achieve those subgoals (Bakker and Schmidhuber 2004; Ring 1994; Schmidhuber and Wahnsiedler 1992; Wiering and Schmidhuber 1998a).”