Flexible shaping: how learning in small steps helps.

By Kai A. Krueger, Peter S. Dayan. 2009

252 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 Artificial Intelligence

Citations in the corpus

Continual Lifelong Learning with Neural Networks: A Review (2018, background): “fers from the highly organized manner in which humans and animals efficiently learn from samples presented in a meaningful order for the shaping of increasingly complex concepts and skills (Krueger and Dayan, 2009; Skinner, 1958).”
Practical Recommendations for Gradient-Based Training of Deep Architectures (2012, background): “There are also results showing that specific ways of setting the initial distribution and ordering of examples (“curriculum learning”) can yield better solutions (Elman, 1993; Bengio et al., 2009; Krueger and Dayan, 2009).”
Knowledge Matters: Importance of Prior Information for Optimization (2013, background): “The idea that learning can be enhanced by guiding the learner through intermediate easier tasks is old, starting with animal training by shaping (Skinner, 1958; Peterson, 2004; Krueger and Dayan, 2009).”
Curriculum learning (2009, background): “Faster convergence with a curriculum was already observed in (Krueger & Dayan, 2009).”
A Survey on Curriculum Learning (2021, background): “This strategy is also revisited in [94] and [52], the latter of which provides evidence for faster convergence.”
Curriculum Learning by Transfer Learning: Theory and Experiments with Deep Networks (2018, background): “Like-wise, the task may be divided by the teacher into smaller sub-tasks, a process sometimes called shaping (Krueger & Dayan, 2009) forcement learning (e.g. Graves et al., 2017).”
On The Power of Curriculum Learning in Training Deep Networks (2019, background): “Imposing a curriculum in order to speed up learning is widely used in the context of human learning and animal training (Skinner, 1958; Pavlov, 2010; Krueger & Dayan, 2009).”
Self-Training for Jointly Learning to Ask and Answer Questions (2018, background): “Curriculum Learning: Studies in cognitive science (Skinner, 1958; Peterson, 2004; Krueger and Dayan, 2009) have shown that humans learn much better when the training examples are not randomly presented but organized in increasing order of difﬁculty.”
Easy Questions First? A Case Study on Curriculum Learning for Question Answering (2016, background): “Studies in cognitive science (Skinner, 1958; Pe-terson, 2004; Krueger and Dayan, 2009) have shown that humans learn much better when the training examples are not randomly presented but organized in increasing order of difﬁculty.”
Neural networks and neuroscience-inspired computer vision. (2014, background): “Even more promising, early work on LSTMs is already feeding back onto neuroscience, having spawned several theories about neural structures responsible for sequence learning involving prefrontal cortex and the basal ganglia [95,96].”
The atoms of neural computation (2014, background): “This work also highlights the relationships between long short-term memory (LSTM) networks (Hochreiter & Schmidhuber, 1997) and models of PFC/basal ganglia functions (Hazy, Frank, & O’Reilly, 2006; Krueger & Dayan, 2009) that include gating and persistent working memory.”
Sub-Task Decomposition Enables Learning in Sequence to Sequence Tasks (2022, background): “The concept of how learning can be enhanced by guiding a learner through intermediate tasks is an old one, dating back to animal training by shaping (Skinner, 1958; Peterson, 2004; Krueger & Dayan, 2009).”
CILEA-NET: Curriculum-Based Incremental Learning Framework for Remote Sensing Image Classification (2021, background): “Moreover, it is apparent from [12] and [13] that animals can learn much faster when a task is decomposed into subtasks from easy to complex based on a particular curriculum.”
Deep Local Trajectory Replanning and Control for Robot Navigation (2019, methodology): “We learn the parameters of the Neural Networks of our system through back-propagation with three Adam optimization procedures [59] aimed at providing an easier path to learning, similar to curriculum learning [60] or shaping [61].”
Simple Substrates for Complex Cognition (2008, background): “However, the sloth of learning may be at least partially ameliorated through structured shaping of competence, itself an underinvestigated topic (Krueger and Dayan, 2007).”
Automatic Curriculum Learning For Deep RL: A Short Survey (2020, methodology): “In the following years, curriculum learning was applied to organize the presentation of training examples or the growth in model capacity in various supervised learning settings [Elman, 1993; Krueger and Dayan, 2009; Bengio et al., 2009].”