The temporal paradox of Hebbian learning and homeostatic plasticity

Zenke, Friedemann; Gerstner, Wulfram; Ganguli, Surya

doi:10.1101/116400

Cited by 155 publications

(250 citation statements)

References 116 publications

Supporting

Mentioning

250

Contrasting

Order By: Relevance

“…Finally, we acknowledge that our pretraining intervention only mitigated, but did not remove, the effect of catastrophic interference. Indeed, it is very likely that other mechanisms, including weight protection schemes that allow new learning to be allocated to synapses according to their importance for past learning, will also play a key role ( 3 , 48 ). A promising avenue for future research may be to understand how structure learning and resource allocation schemes interact.…”

Section: Discussionmentioning

confidence: 99%

Comparing continual task learning in minds and machines

Flesch

Balaguer

Dekker

et al. 2018

Proc. Natl. Acad. Sci. U.S.A.

125

136

View full text Add to dashboard Cite

SignificanceHumans learn to perform many different tasks over the lifespan, such as speaking both French and Spanish. The brain has to represent task information without mutual interference. In machine learning, this “continual learning” is a major unsolved challenge. Here, we studied the patterns of errors made by humans and state-of-the-art neural networks while they learned new tasks from scratch and without instruction. Humans, but not machines, seem to benefit from training regimes that blocked one task at a time, especially when they had a prior bias to represent stimuli in a way that encouraged task separation. Machines trained to exhibit the same prior bias suffered less interference between tasks, suggesting new avenues for solving continual learning in artificial systems.

show abstract

Section: Discussionmentioning

confidence: 99%

Comparing continual task learning in minds and machines

Flesch

Balaguer

Dekker

et al. 2018

Proc. Natl. Acad. Sci. U.S.A.

125

136

View full text Add to dashboard Cite

show abstract

“…Unlike biological odor learning, artificial neural networks optimized for a certain task tend to suffer from catastrophic forgetting, and the pursuit of online learning capabilities in deep networks is a subject of active study (McCloskey and Cohen, 1989; Kemker and Kanan, 2017; Kirkpatrick et al, 2017; Velez and Clune, 2017; Zenke et al, 2017; Serrà et al, 2018). In contrast, the EPLff learning network described herein naturally resists catastrophic forgetting, exhibiting powerful online learning using a fast spike timing-based coding metric.…”

Section: Resultsmentioning

confidence: 99%

A Spike Time-Dependent Online Learning Algorithm Derived From Biological Olfaction

Borthakur

Cleland

2019

Front. Neurosci.

View full text Add to dashboard Cite

We have developed a spiking neural network (SNN) algorithm for signal restoration and identification based on principles extracted from the mammalian olfactory system and broadly applicable to input from arbitrary sensor arrays. For interpretability and development purposes, we here examine the properties of its initial feedforward projection. Like the full algorithm, this feedforward component is fully spike timing-based, and utilizes online learning based on local synaptic rules such as spike timing-dependent plasticity (STDP). Using an intermediate metric to assess the properties of this initial projection, the feedforward network exhibits high classification performance after few-shot learning without catastrophic forgetting, and includes a none of the above outcome to reflect classifier confidence. We demonstrate online learning performance using a publicly available machine olfaction dataset with challenges including relatively small training sets, variable stimulus concentrations, and 3 years of sensor drift.

show abstract

“…The higher importance value denotes the importance of a particular parameter for the previous tasks. Similarly, in the synaptic intelligence (SI) approach, the importance value for each parameter is assigned according to the change in the parameter values [32]. In both the above approaches (EWC and SI), the importance is given for each synapse, while in the proposed model, the importance is given for the direction at the neuronal level.…”

Section: Comparison To Other Modelsmentioning

confidence: 99%

“…This approach keeps a copy of the diagonals of the Fisher information matrix and the latest learned parameters. Zenke et al proposed the Synaptic Intelligence model, where the importance of each parameter for the previous tasks is estimated based on the change in parameter values [32]. Based on this value, the parameters with high importance refrain from learning, and the rest of the parameters are allowed to update.…”

Section: Introductionmentioning

confidence: 99%

Memory Consolidation with Orthogonal Gradients for avoiding Catastrophic Forgetting

Kanagamani

Krishnamurthy

Chakravarthy

et al. 2022

Preprint

View full text Add to dashboard Cite

The memory consolidation process enables the accumulation of recent and remote memories in the long-term memory store. In general, the deep network models of memory suffer from forgetting old information while learning new information, called catastrophic forgetting/interference. The human brain overcomes this problem quite effectively, a problem that continues to challenge current deep neural network models. We propose a regularization-based model to solve the problem of catastrophic forgetting. According to the proposed training mechanism, the network parameters are constrained to vary in a direction orthogonal to the average of the error gradients corresponding to the previous tasks. We also ensure that the constraint used in parameter updating satisfies the locality principle. The proposed model's performance is compared with Elastic Weight Consolidation on standard datasets such as permuted MNIST and split MNIST on classification tasks using fully connected networks, and Convolution-based networks. The model performance is also compared to an autoencoder on split MNIST dataset, and to complex core50 dataset on two types of classification tasks with EWC. The proposed model gives a new view on plasticity at the neuronal level. In the proposed model, the parameter updating is controlled by the neuronal level plasticity rather than synapse level plasticity as in other standard models. The biological plausibility of the proposed model is discussed by linking the extra parameters to synaptic tagging, which represents the state of the synapse involved in Long Term Potentiation (LTP).

show abstract

The temporal paradox of Hebbian learning and homeostatic plasticity

Cited by 155 publications

References 116 publications

Comparing continual task learning in minds and machines

Comparing continual task learning in minds and machines

A Spike Time-Dependent Online Learning Algorithm Derived From Biological Olfaction

Memory Consolidation with Orthogonal Gradients for avoiding Catastrophic Forgetting

Contact Info

Product

Resources

About