Catastrophic Forgetting in Connectionist Networks: Causes, Consequences and Solutions

By R. M. French. 1994

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

Structural Synaptic Plasticity Has High Memory Capacity and Can Explain Graded Amnesia, Catastrophic Forgetting, and the Spacing Effect (2014, background): “ing catastrophic forgetting in brains [36,37], graded retrograde”
Maximum a posteriori adaptation of network parameters in deep models (2015, background): “Such a phenomenon is commonly referred to as catastrophic forgetting [28].”
Ensemble based systems in decision making (2006, background): “This approach, however, results in loss of all previously acquired information, a phenomenon known as catastrophic forgetting (or interfering) [74].”
Developmental Plasticity in Cartesian Genetic Programming based Neural Networks (2016, background): “Catastrophic forgetting (CF) is the phenomenon in which a trained artificial neural network loses its accuracy when the network is trained again on a different computational problem (McCloskey and Cohen, 1989; Ratcliff, 1990; Sharkey and Sharkey, 1995; Ans et al., 2002; French, 1994).”
In search of intelligence: evolving a developmental neuron capable of learning (2014, background): “…is to devise a neural developmental model capable of producing neural networks that can solve multiple problems (i.e. not suffer from a phenomenon called catastrophic interference; Ans, Rousset, French, & Musca, 2002; French, 1994; McCloskey & Cohen, 1989; Ratcliff, 1990; Sharkey & Sharkey, 1995).”
BRINGING UP ROBOT: FUNDAMENTAL MECHANISMS FOR CREATING A SELF-MOTIVATED, SELF-ORGANIZING ARCHITECTURE (2005, background): “One serious problem that can occur has been termed catastrophic forgetting [5], which occurs when new information completely destroys previously learned information.”
Neuro-Centric and Holocentric Approaches to the Evolution of Developmental Neural Networks (2014, unknown): “SD gives rise to “catastrophic forgetting” (CF) [3–5].”
Sequential learning in neural networks: A review and a discussion of pseudorehearsal based methods (2004, unknown): “French [19] also proposed a dual network strategy, but his “pseudorecurrent” architecture has shared input units at the ﬁrst layer, as illustrated in Fig.”
Learning hidden unit contributions for unsupervised speaker adaptation of neural network acoustic models (2014, background): “This formulation has several advantages: first, the total number of adaptation parameters stays relatively low - at most the total number of hidden units, ∑L l M ; second, since it does not make assumptions about the form of nonlinearity, neither the internal structure of a layer, it may be applied ”