Related Experiment Videos
A working memory model based on fast Hebbian learning.
A Sandberg1, J Tegnér, A Lansner
1Department of Numerical Analysis and Computer Science, Royal Institute of Technology, 100 44 Stockholm, Sweden. asa@nada.kth.se
Network (Bristol, England)
|December 5, 2003
Summary
Working memory may rely on fast Hebbian synaptic plasticity, not just persistent activity. This new model offers greater resistance to network errors and supports storing multiple memories.
Area of Science:
- Computational neuroscience
- Cognitive neuroscience
Background:
- Working memory (WM) is crucial for cognition.
- Current models often propose WM relies on persistent neural activity ('bump' states) maintained by specific synaptic weights.
- These models face challenges with network stability and distractor resistance.
Purpose of the Study:
- To propose and test an alternative hypothesis for working memory.
- To investigate the role of fast Hebbian synaptic plasticity in working memory.
- To develop a computational model that overcomes limitations of previous WM models.
Main Methods:
- Developed a computational model based on fast Hebbian synaptic plasticity.
- Simulated the oculomotor delayed response task.
- Assessed model performance regarding distractor resistance and network inhomogeneity.
Main Results:
- The Hebbian plasticity model demonstrated robust working memory function.
- The model showed enhanced resistance to distractors compared to 'bump' state models.
- The model exhibited greater tolerance to network inhomogeneity.
- The model successfully stored multiple memories.
Conclusions:
- Fast Hebbian synaptic plasticity provides a viable and potentially more robust mechanism for working memory.
- This plasticity-based model offers advantages in stability and capacity over persistent activity models.
- Future research should explore the biological plausibility and experimental validation of this mechanism.