Latching dynamics in neural networks with synaptic depression

Area of Science:

• Computational neuroscience
• Cognitive psychology
• Neural networks

Background:

• Prediction involves activating target concepts from prime stimuli.
• Prime-target relations are linked to long-term memory formation.
• Hopfield networks with Hebbian connectivity model long-term memories.

Purpose of the Study:

• Investigate the conditions for latching dynamics in priming models.
• Analyze the role of short-term synaptic depression in memory processing.
• Evaluate the robustness and biological realism of latching dynamics models.

Main Methods:

• Analytical and numerical approaches.
• Modeling latching dynamics in Hopfield networks.
• Examining symmetric Hebbian learning rules.

Main Results:

• Latching dynamics can exist with symmetric Hebbian learning.
• The dynamics lack robustness and impose biologically unrealistic constraints.
• Symmetry of the Hebbian rule does not prevent latching dynamics, but parameter fine-tuning is necessary.

Conclusions:

• Latching dynamics offer a computational model for priming.
• Symmetric Hebbian learning can support latching dynamics.
• Further refinement is needed to enhance model realism and robustness.