Dynamic control of sequential retrieval speed in networks with heterogeneous learning rules.

Area of Science:

• Computational neuroscience
• Neural dynamics
• Learning and memory

Background:

• Sequential neural activity is temporally rescaled in brain areas during time estimation and variable-speed motor tasks.
• Existing network models using Hebbian learning rules can learn and retrieve sequences, but typically at a fixed speed.

Purpose of the Study:

• To investigate how heterogeneity in neural plasticity rules affects network dynamics.
• To explore mechanisms for controlling the speed of sequential neural activity retrieval.

Main Methods:

• Developed a computational model where neurons exhibit varying degrees of temporal symmetry in their plasticity rules.
• Analyzed network dynamics under different external input conditions.

Main Results:

• Network retrieval speed is controllable by external inputs.
• Temporally symmetric plasticity rules act as 'brakes,' slowing dynamics.
• Temporally asymmetric rules act as 'accelerators,' speeding up dynamics.
• Networks can generate distinct preparatory and execution activity patterns.

Conclusions:

• Heterogeneity in neural plasticity rules provides a mechanism for flexible temporal rescaling of sequential neural activity.
• This model offers insights into how the brain controls timing during complex behaviors and generates distinct activity states.