Models in NEST¶

What we mean by models¶

The term models in the context of NEST (and the field of computational neuroscience as a whole) is used with two different meanings:

Neuron and synapse models. These consist of a set of mathematical equations and algorithmic components that describe the characteristics and behavior of biological neurons and synapses. In NEST, the terms neuron and synapse models are also used for the C++ implementations of these conceptual entities. Most of the models in NEST are based on either peer-reviewed publications or text books like [1].
Network models. These models are created from individual neuron and synapse models using the different commands provided by the PyNEST API. Examples for such network models are the microcircuit model or the multi-area model). In the following description, we focus on neuron and synapse models and not on network models.

Find a model¶

By default, NEST comes with a ton of models! Textbook standards like integrate-and-fire and Hodgkin-Huxley-type models are available alongside high-quality implementations of models published by the neuroscience community. The model directory is organized by keywords (e.g., adaptive threshold, conductance-based, etc.). Models that contain a specific keyword will be listed under that word.

In many modeling situations, the full set of models that ship with NEST by default is not needed. To only include a subset of the models with NEST, please have a look at the modelset configuration options.

Create and customize models with NESTML¶

Check out NESTML, a domain-specific language for neuron and synapse models. NESTML enables fast prototyping of new models using an easy to understand, yet powerful syntax. This is achieved by a combination of a flexible processing toolchain written in Python with high simulation performance through the automated generation of C++ code, suitable for use in NEST Simulator.

Model naming¶

Neuron models¶

Neuron model names in NEST combine abbreviations that describe the dynamics and synapse specifications for that model. They may also include the author’s name of a model based on a specific paper.

For example, the neuron model name

iaf_cond_beta

corresponds to an implementation of a spiking neuron using integrate-and-fire dynamics with conductance-based synapses. Incoming spike events induce a postsynaptic change of conductance modeled by a beta function.

As an example for a neuron model name based on specific paper,

hh_cond_exp_traub

implements a modified version of the Hodgkin Huxley neuron model based on Traub and Miles (1991)

Synapse models¶

Synapse models include the word synapse as the last word in the model name.

Synapse models may begin with the author name (e.g., clopath_synapse) or process (e.g., stdp_synapse).

Devices¶

A device name should represent its physical counterpart - like a multimeter is multimeter. In general, the term recorder is used for devices that store the output (e.g., spike times or synaptic strengths over time) of other nodes and make it accessible to the user. The term generator is used for devices that provide input into the simulation.