Auto-teaching: networks that develop their own teaching input

By Stefano Nolfi, Domenico Parisi. 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

Born to Learn: the Inspiration, Progress, and Future of Evolved Plastic Artificial Neural Networks (2017, background): “The ideas of evolving learning was implemented in Nolﬁ and Parisi (1993) by evolving “auto-teaching” inputs, which could then provide an error signal for the network to adjust weights during lifetime.”
Evolving plastic neural networks with novelty search (2010, background): “For example, Nolfi and Parisi (1993, 1996) evolved self-teaching networks that trained a motor control network through backpropagation from the outputs of a teaching subnetwork.”
Learning and Evolution in Neural Networks (1994, background): “…in which the teaching input for backpropagation learning is generated by the same network that learns on the basis of this teaching input, cf. Nolfi & Parisi, 1993) Hence, we should choose a learning task for our networks for which the assumption of such external "teacher" is a plausible one.”
Evolving adaptive neural networks with and without adaptive synapses (2003, background): “Nolfi and Parisi used backpropagation inside evolved networks to make predictions about future states (Nolfi et al. 1990, 1994) and also for a network to train itself (Nolfi and Parisi 1993, 1995).”
Learning to Adapt to Changing Environments in Evolving Neural Networks (1996, methodology): “The creatures have the possibility to learn using the backpropagation procedure but, since we did not want to provide them with the correct teaching input at each time step as decided by the researcher, we used a self-teaching architecture (cf. Nolfi and Parisi, 1993).”
Learning and Evolution (1999, background): “This explains the surprising result described in section 6 that self-generated teaching inputs do not correspond to the desired motor actions although they are capable of producing changes that generate suitable motor actions (on this point see also Nolfi and Parisi, 1993).”