Introducing the Dendrify framework for incorporating dendrites to spiking neural networks

Area of Science:

• Computational neuroscience
• Neural network modeling

Background:

• Understanding subcellular neuronal features is key to circuit processing.
• The role of dendritic computations in network operations is largely unexplored due to limitations in current modeling tools.

Purpose of the Study:

• To bridge the gap between simplistic spiking neural networks and computationally expensive detailed models.
• To introduce Dendrify, an open-source Python package for generating reduced compartmental neuron models.

Main Methods:

• Dendrify automatically generates reduced compartmental neuron models using simple commands.
• The package is based on the Brian 2 simulator.
• Models incorporate simplified yet biologically relevant dendritic and synaptic integrative properties.

Main Results:

• Dendrify provides a balance between flexibility, performance, and biological accuracy.
• Enables exploration of dendritic contributions to network-level functions.
• Facilitates the development of more powerful neuromorphic systems.

Conclusions:

• Dendrify allows for realistic and efficient network models that account for dendritic properties.
• The tool supports the adoption of dendritic features in spiking neural networks.
• It aids in understanding the impact of dendritic computations on neural circuit function.