Warning

This is A PREVIEW for NEST 3.0 and NOT an OFFICIAL RELEASE! Some functionality may not be available and information may be incomplete!

# NEST 3.0: What’s new?¶

## Introduction¶

NEST 3.0 introduces a more direct approach to accessing node properties and handling connections. The changes will allow you to perform operations that were not possible in previous versions.

Python 3

With NEST 3.0, we no longer support Python 2. Running the code snippets throughout this guide requires a freshly started instance of Python (and sometimes pyplot from matplotlib). Check out our Installation instructions for more information on the dependencies.

See our NEST 3.0: Detailed transition guide to see a full list of functions that have changed.

## What’s new?¶

### New functionality for node handles (neurons and devices)¶

In NEST 3.0, nest.Create() returns a NodeCollection object instead of a list of global IDs. This provides a more compact and flexible way for handling nodes.

In most use cases, you will not need to make any changes to your scripts in NEST 3.0, unless you have used topology or subnets.

NodeCollection supports the following functionality:

NEST 2.x

NEST 3.0

# A list of 10 GIDs is returned
nrns = nest.Create('iaf_psc_alpha', 10)

# Use lists as arguments in Connect
nest.Connect(nrns, nrns)

# A NodeCollection object is returned
nrns = nest.Create('iaf_psc_alpha', 10)

# Use NodeCollection objects as
# arguments in Connect
nest.Connect(nrns, nrns)


NodeCollections support the following operations:

Printing

A compact representation of information about the NodeCollection can be printed

>>>  nrns = nest.Create('iaf_psc_alpha', 10)
>>>  print(nrns)

Indexing

Indexing returns a new NodeCollection with a single node

>>>  print(nrns[3])


NodeCollections support array indexing. Array indexing is done by passing a list or tuple of indices when indexing. A NodeCollection with the node IDs at the chosen indices is then returned. Note that all indices must be strictly ascending and unique because all node IDs in a NodeCollection must be unique.

>>>  print(nrns[[1, 2, 5, 6]])
model=iaf_psc_alpha, size=2, first=2, last=3;
model=iaf_psc_alpha, size=2, first=6, last=7)


One may also pass a list or tuple of Booleans, where the returned NodeCollection contains the True elements of the list or tuple. The length of the list of tuple of Booleans must be equal to the length of the NodeCollection.

>>>  print(nrns[[True, True, True, True, False, False, True, True, True, True]])
model=iaf_psc_alpha, size=4, first=1, last=4;
model=iaf_psc_alpha, size=4, first=7, last=10)

Iteration

You can iterate the nodes in a NodeCollection and receive a single element NodeCollection

>>>   for node in nrns:
>>>       print(node.global_id)
1
2
3
4
5
6
7
8
9
10

Slicing

A NodeCollection can be sliced in the same way one would slice a list, with start:stop:step inside brackets

>>>  print(nrns[2:9:3])
model=iaf_psc_alpha, size=2, first=3, last=9, step=3)

Getting the size

You can easily get the number of nodes in the NodeCollection with

>>>  len(nrns)
10

Conversion to and from lists

NodeCollections can be converted to lists of node IDs

>>>  nrns.tolist()
[1, 2, 3, 4, 5, 6, 7, 8, 9, 10]


And you can create a NodeCollection by providing a list, tuple, NumPy array or range of node IDs

>>>  print(nest.NodeCollection([2, 3, 4, 8]))
model=iaf_psc_alpha, size=3, first=2, last=4;
model=iaf_psc_alpha, size=1, first=8)
>>>  print(nest.NodeCollection(range(1,4)))


Note however that the nodes have to be already created. If any of the node IDs refer to a non existing node, an error is thrown. Additionally each node ID can only occur once and the list of node IDs must be sorted in ascending order.

Composing

When composing two NodeCollections, NEST tries to concatenate the two into a single NodeCollection.

>>>  nrns = nest.Create('iaf_psc_alpha', 10)
>>>  nrns_2 = nest.Create('iaf_psc_alpha', 3)
>>>  print(nrns + nrns_2)


If the node IDs are not continuous or the models are different, a composite will be created:

>>>  nrns_3 = nest.Create('iaf_psc_delta', 3)
>>>  print(nrns + nrns_3)
model=iaf_psc_alpha, size=10, first=1, last=10;
model=iaf_psc_delta, size=3, first=14, last=16)


Note that composing NodeCollections that overlap or that contain metadata (see section on spatially distributed nodes) is not supported.

Test of equality

You can test if two NodeCollections are equal, i.e. that they contain the same node IDs

>>>  nrns == nrns_2
False
>>>  nrns_2 == nest.NodeCollection([11, 12, 13])
True

Test of membership

You can test if a NodeCollection contains a certain ID

>>>  2 in nrns
True
>>>  11 in nrns
False

Direct attributes

You can directly get and set parameters of your NodeCollection

>>> nrns.V_m = [-70., -60., -50., -40., -30., -20., -10., -20., -30., -40.]
>>> nrns.V_m
(-70.0, -60.0, -50.0, -40.0, -30.0, -20.0, -10.0, -20.0, -30.0, -40.0)
>>> nrns.C_m = 111.
>>> nrns.C_m
(111.0, 111.0, 111.0, 111.0, 111.0, 111.0, 111.0, 111.0, 111.0, 111.0)


If your nodes are spatially distributed (see spatially distributed nodes), you can also get the spatial properties of the nodes

>>> spatial_nodes.spatial
{'center': (0.0, 0.0),
'edge_wrap': False,
'extent': (1.0, 1.0),
'network_size': 4,
'shape': (2, 2)}


### Get the node status¶

get() returns the parameters in the collection. You can call get() in several ways.

To get all parameters in the collection, use get() without any function arguments. This returns a dictionary with tuples.

>>>    nodes_exp = nest.Create('iaf_psc_exp', 5)
>>>    nodes_exp[:3].get()
{'archiver_length': (0, 0, 0),
'beta_Ca': (0.001, 0.001, 0.001),
'C_m': (250.0, 250.0, 250.0),
'Ca': (0.0, 0.0, 0.0),
'delta': (0.0, 0.0, 0.0),
'E_L': (-70.0, -70.0, -70.0),
'element_type': ('neuron', 'neuron', 'neuron'),
'frozen': (False, False, False),
'global_id': (11, 12, 13),
'I_e': (0.0, 0.0, 0.0),
'local': (True, True, True),
'model': ('iaf_psc_exp', 'iaf_psc_exp', 'iaf_psc_exp'),
'node_uses_wfr': (False, False, False),
'post_trace': (0.0, 0.0, 0.0),
'recordables': (('I_syn_ex',
'I_syn_in',
'V_m',
'weighted_spikes_ex',
'weighted_spikes_in'),
('I_syn_ex', 'I_syn_in', 'V_m', 'weighted_spikes_ex', 'weighted_spikes_in'),
('I_syn_ex', 'I_syn_in', 'V_m', 'weighted_spikes_ex', 'weighted_spikes_in')),
'rho': (0.01, 0.01, 0.01),
'supports_precise_spikes': (False, False, False),
'synaptic_elements': ({}, {}, {}),
't_ref': (2.0, 2.0, 2.0),
't_spike': (-1.0, -1.0, -1.0),
'tau_Ca': (10000.0, 10000.0, 10000.0),
'tau_m': (10.0, 10.0, 10.0),
'tau_minus': (20.0, 20.0, 20.0),
'tau_minus_triplet': (110.0, 110.0, 110.0),
'tau_syn_ex': (2.0, 2.0, 2.0),
'tau_syn_in': (2.0, 2.0, 2.0),
'V_m': (-70.0, -70.0, -70.0),
'V_reset': (-70.0, -70.0, -70.0),
'V_th': (-55.0, -55.0, -55.0),
'vp': (0, 0, 0)}


To get specific parameters in the collection, use get([parameter_name_1, parameter_name_2, ... , parameter_name_n]).

Get the parameters V_m and V_reset of all nodes

>>>    nodes = nest.Create('iaf_psc_alpha', 10, {'V_m': -55.})
>>>    nodes.get(['V_m', 'V_reset'])
{'V_m': (-55.0, -55.0, -55.0, -55.0, -55.0, -55.0, -55.0, -55.0, -55.0, -55.0),
'V_reset': (-70.0,
-70.0,
-70.0,
-70.0,
-70.0,
-70.0,
-70.0,
-70.0,
-70.0,
-70.0)}


To get a specific parameter from the collection, you can use get(parameter_name). This will return a tuple with the values of that parameter.

>>>    nodes.get('t_ref')
(2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0, 2.0)


If you have a single-node NodeCollection, get() will return a dictionary with single values or a single value, depending on how it is called.

>>>    nodes[0].get(['V_m', 'V_reset'])
{'V_m': -55.0, 'V_reset': -70.0}
>>>    nodes[0].get('t_ref')
2.0


To select fields at a deeper hierarchy level, use get(parameter_name, property_name), this will return an array. You can also use get(parameter_name, [property_name_1, ..., property_name_n]) and get a dictionary with arrays.

>>>    sr = nest.Create('spike_recorder')
>>>    sr.get('events', 'senders')
array([], dtype=int64)


Lastly, you can specify the output format (pandas and JSON for now). The output format can be specified for all the different get() versions above.

>>>    nodes[0].get(['V_m', 'V_reset'], output='json')
'{"V_m": -55.0, "V_reset": -70.0}'


### Set node properties¶

set() sets the values of a parameter by iterating over each node.

As with get(), you can set parameters in different ways.

To set several parameters at once, use nodes.set(parameter_dict), where the keys of the parameter_dict are the parameter names. The values could be a list the size of the NodeCollection, a single value, or a nest.Parameter.

nodes[:3].set({'V_m': [-70., -80., -90.], 'C_m': 333.})


You could also set a single parameter by using nodes.set(parameter_name=parameter). As parameter, you can either send in a single value, a list the size of the NodeCollection, or a nest.Parameter

nodes.set(t_ref=3.0)
nodes[:3].set(t_ref=[3.0, 4.0, 5.0])
nodes.set(t_ref=nest.random.uniform())


Note that some parameters, like global_id, cannot be set. The documentation of a specific model will point out which parameters can be set and which are read-only.

### Dictionary with lists when setting parameters¶

It is now possible to use a dictionary with lists when setting node parameters with Create(), set() or SetStatus(). The values of the lists will be distributed across the nodes. The way to do this previously was to apply a list of dictionaries. This is still possible.

The values in the single dictionary can also be single values; the value will then be applied to each node. You can mix and match as you want; the dictionary can contain lists and single values at the same time.

parameter_list = {"I_e": [200.0, 150.0], "tau_m": 20.0, "V_m": [-77.0, -66.0]}
pop = nest.Create("iaf_psc_alpha", 2, params= {"I_e": [200.0, 150.0], "tau_m": 20.0, "V_m": [-77.0, -66.0]})

print(pop.get(["I_e", "tau_m", "V_m"]))


### New functionality for connecting arrays of node IDs¶

While you should aim to use NodeCollections to create connections whenever possible, there may be cases where you have a predefined set of pairs of pre- and postsynaptic nodes. In those cases, it may be inefficient to convert the individual IDs in the pair to NodeCollections to be passed to the Connect() function, especially if there are thousands or millions of pairs to connect.

To efficiently create connections in these cases, you can pass NumPy arrays to Connect(). This variant of Connect() will create connections in a one-to-one fashion.

nest.Create('iaf_psc_alpha', 10)
# Node IDs in the arrays must address existing nodes, but may occur multiple times.
sources = np.array([1, 5, 7, 5], dtype=np.uint64)
targets = np.array([2, 2, 4, 4], dtype=np.uint64)
nest.Connect(sources, targets, conn_spec="one_to_one")


You can also specify weights, delays, and receptor type for each connection as arrays. All arrays have to have lengths equal to those of sources and targets.

weights = np.array([0.5, 0.5, 2., 2.])
delays = np.array([1., 1., 2., 2.])
syn_spec = {'weight': weights, 'delay': delays}
nest.Connect(sources, targets, conn_spec='one_to_one', syn_spec=syn_spec)


### New functionality for handling connections (synapses)¶

Just like a NodeCollection is a container for node IDs, a SynapseCollection is a container for connections. In NEST 3, when you call GetConnections() a SynapseCollection is returned. SynapseCollections support a lot of the same operations as NodeCollections.

SynapseCollection supports:

You can find a full example in our example network page.

Printing

Printing a SynapseCollection produces a table source and target node IDs, synapse model, weight and delay. If your SynapseCollection has more than 36 elements, only the first and last 15 connections are displayed. To print all, first set print_all = True on your SynapseCollection.

>>>  nest.Connect(nodes[:2], nodes[:2])
>>>  synColl = nest.GetConnections()
>>>  print(synColl)
source   target   synapse model   weight   delay
-------- -------- --------------- -------- -------
1        1  static_synapse    1.000   1.000
1        2  static_synapse    1.000   1.000
2        1  static_synapse    1.000   1.000
2        2  static_synapse    1.000   1.000

>>> synColl.print_all = True

Indexing

Indexing returns a single connection SynapseCollection.

>>>  print(synColl[1])
source   target   synapse model   weight   delay
-------- -------- --------------- -------- -------
1        2  static_synapse    1.000   1.000

Iteration

A SynapseCollection can be iterated, yielding a single connection SynapseCollections.

>>>  for conn in synColl:
>>>      print(conn.source)
1
1
2
2

Slicing

A SynapseCollection can be sliced with start:stop:step inside brackets

>>>  print(synColl[0:3:2])
source   target   synapse model   weight   delay
-------- -------- --------------- -------- -------
1        1  static_synapse    1.000   1.000
2        1  static_synapse    1.000   1.000

Getting the size

We can get the number of connections in the SynapseCollection with

>>>  len(synColl)
4

Test of equality

Two SynapseCollections can be tested for equality, i.e. that they contain the same connections.

>>>  synColl == synColl
True
>>>  synColl[:2] == synColl[2:]
False

Getting connection parameters

Just as with NodeCollection, you can get parameters of the connections with get(). The same function arguments as for NodeCollections get() apply here. The returned values also follow the same rules.

If you call get() without any arguments, a dictionary with all parameters is returned as a list if the number of connections is bigger than 1 and a single integer if number of connections is equal to 1.

>>>  synColl.get()
{'delay': [1.0, 1.0, 1.0, 1.0],
'port': [0, 1, 2, 3],
'receptor': [0, 0, 0, 0],
'sizeof': [32, 32, 32, 32],
'source': [1, 1, 2, 2],
'synapse_id': [0, 0, 0, 0],
'synapse_model': ['static_synapse',
'static_synapse',
'static_synapse',
'static_synapse'],
'target': [1, 2, 1, 2],
'weight': [1.0, 1.0, 1.0, 1.0]}


Calling get(parameter_name) will return a list of parameter values, while get([parameter_name_1, ... , parameter_name_n]) returns a dictionary with the values.

>>>  synColl.get('weight')
[1.0, 1.0, 1.0, 1.0]

>>>  synColl[2].get(['source', 'target'])
{'source': 2, 'target': 1}


It is also possible to select an alternative output format with the output keyword. Currently it is possible to get the output in a json format, or a Pandas format (if Pandas is installed).

Setting connection parameters

Likewise, you can set the parameters of connections in the SynapseCollection. Again the same rules as with set() on NodeCollection applies, see Set node properties for more details.

If you want to set several parameters at once, use set(parameter_dictionary). You can use a single value, a list, or a nest.Parameter as values. If a single value is given, the value is set on all connections.

>>>  synColl.set({'weight': [1.5, 2.0, 2.5, 3.0], 'delay': 2.0})


Updating a single parameter is done by calling set(parameter_name=parameter_value). Again you can use a single value, a list, or a nest.Parameter as value.

>>>  synColl.set(weight=3.7)

>>>  synColl.set(weight=[4.0, 4.5, 5.0, 5.5])


Note that some parameters, like source and target, cannot be set. The documentation of a specific model will point out which parameters can be set and which are read-only.

Setting and getting attributes directly

You can also directly get and set parameters of your SynapseCollection

>>>  synColl.weight = 5.
>>>  synColl.weight
[5.0, 5.0, 5.0, 5.0]
>>>  synColl.delay = [5.1, 5.2, 5.3, 5.4]
>>>  synColl.delay
[5.1, 5.2, 5.3, 5.4]


If you use a list to set the parameter, the list needs to be the same length as the SynapseCollection.

For spatially distributed sources and targets, you can access the distance between the source-target pairs by calling distance on your SynapseCollection.

>>>  synColl.distance
(0.47140452079103173,
0.33333333333333337,
0.4714045207910317,
0.33333333333333337,
3.925231146709438e-17,
0.33333333333333326,
0.4714045207910317,
0.33333333333333326,
0.47140452079103157)

Iterator of sources and targets

Calling SynapseCollection.sources() or SynapseCollection.targets() returns an iterator over the source IDs or target IDs, respectively.

>>>  print([s*3 for s in synColl.sources()])
[3, 3, 6, 6]


### Collocated synapses¶

It is now possible to create connections with several synapses simultaneously. The different synapse dictionaries will then be applied to each source-target pair. To create these collocated synapses, CollocatedSynapses() must be used as the syn_spec argument of Connect, instead of the usual syn_spec dictionary argument. CollocatedSynapses() takes dictionaries as arguments.

nodes = nest.Create('iaf_psc_alpha', 3)
syn_spec = nest.CollocatedSynapses({'weight': 4., 'delay': 1.5},
{'synapse_model': 'stdp_synapse'},
{'synapse_model': 'stdp_synapse', 'alpha': 3.})
nest.Connect(nodes, nodes, conn_spec='one_to_one', syn_spec=syn_spec)

conns = nest.GetConnections()
print(conns.alpha)


This will create 9 connections: 3 using static_synapse with a weight of 4 and delay of 1.5, and 6 using the stdp_synapse. Of the 6 using stdp_synapse, 3 will have the default alpha value, and 3 will have an alpha of 3.0.

>>> print(nest.GetKernelStatus('num_connections'))
9


If you want to connect with different receptor types, you can do the following:

src = nest.Create('iaf_psc_exp_multisynapse', 7)
trgt = nest.Create('iaf_psc_exp_multisynapse', 7, {'tau_syn': [0.1 + i for i in range(7)]})

syn_spec = nest.CollocatedSynapses({'weight': 5.0, 'receptor_type': 2},
{'weight': 1.5, 'receptor_type': 7})

nest.Connect(src, trgt, 'one_to_one', syn_spec=syn_spec)

conns = nest.GetConnections()
print(conns.get())


You can see how many synapse parameters you have by doing len() on your CollocatedSynapses object:

>>> len(syn_spec)
2


### Parametrization¶

NEST 3.0 introduces parameter objects, i.e., objects that represent values drawn from a random distribution or values based on various spatial node parameters. Parameters can be used to set node status, to create positions in space (see Topology section below), and to define connection probabilities, weights and delays. The parameters can be combined in different ways, and they can be used with some mathematical functions that are provided by NEST.

The following parameters and functionalities are provided:

#### Random parameters¶

The random module contains random distributions that can be used to set node and connection parameters, as well as positions for spatially distributed nodes.

Parameter

Description

nest.random.uniform(min=0.0, max=1.0)


Draws samples based on a uniform distribution.

nest.random.normal(mean=0.0, std=1.0)


Draws samples based on a normal distribution.

nest.random.exponential(beta=1.0)


Draws samples based on a exponential distribution.

nest.random.lognormal(mean=0.0, std=1.0)


Draws samples based on a lognormal distribution.

For every value to be generated, samples are drawn from a distribution. The distribution uses NEST’s random number generator, and are therefore thread-safe. Note that arguments can be passed to each of the distributions above to control the parameters of the distribution.

n = nest.Create('iaf_psc_alpha', 10000, {'V_m': nest.random.normal(mean=-60.0, std=10.0)})

node_ids = n.global_id
v_m = n.get('V_m')
fig, ax = pyplot.subplots(figsize=(12, 6),
gridspec_kw={'width_ratios':
[3, 1]},
ncols=2,
sharey=True)
ax[0].plot(node_ids, v_m, '.', alpha=0.5, ms=3.5)
ax[0].set_xlabel('Node_ID');
ax[1].hist(v_m, bins=40, orientation='horizontal');
ax[1].set_xlabel('num. nodes');
ax[0].set_ylabel('V_m');


#### Spatial parameters¶

The spatial module contains parameters related to spatial positions of the nodes.

To create spatially distributed nodes (see section on spatially distributed nodes for more), use nest.spatial.grid() or nest.spatial.free.

Parameter

Description

nest.spatial.grid(shape, center=None,
extent=None, edge_wrap=False)


Create spatially positioned nodes distributed on a grid with dimensions given by shape=[nx, ny(, nz)].

nest.spatial.free(pos, extent=None,
edge_wrap=False, num_dimensions=None)


Create spatially positioned nodes distributed freely in space with dimensions given by pos or num_dimensions.

grid_nodes = nest.Create('iaf_psc_alpha', positions=nest.spatial.grid(shape=[10, 8]))
nest.PlotLayer(grid_nodes);

free_nodes = nest.Create('iaf_psc_alpha', 100,
positions=nest.spatial.free(nest.random.uniform(min=0., max=10.),
num_dimensions=2))
nest.PlotLayer(free_nodes);


After you have created your spatially distributed nodes, you can use spatial property to set node or connection parameters.

Parameter

Description

nest.spatial.pos.x
nest.spatial.pos.y
nest.spatial.pos.z

Position of a neuron, on the x, y, and z axis.
Can be used to set node properties, but not for connecting.
nest.spatial.source_pos.x
nest.spatial.source_pos.y
nest.spatial.source_pos.z

Position of the source neuron, on the x, y, and z axis.
Can only be used when connecting.
nest.spatial.target_pos.x
nest.spatial.target_pos.y
nest.spatial.target_pos.z

Position of the target neuron, on the x, y, and z axis.
Can only be used when connecting.
nest.spatial.distance

Distance between two nodes. Can only be used when connecting.
nest.spatial.distance.x
nest.spatial.distance.y
nest.spatial.distance.z

Distance on the x, y and z axis between the source and target neuron.
Can only be used when connecting.

These parameters represent positions of neurons or distances between two neurons. To set node parameters, only the node position can be used. The others can only be used when connecting.

positions = nest.spatial.free([[x, 0.5*x] for x in np.linspace(0, 1.0, 10000)])
spatial_nodes = nest.Create('iaf_psc_alpha', positions=positions)

parameter = -60 + nest.spatial.pos.x + (0.4 * nest.spatial.pos.x * nest.random.normal())
spatial_nodes.set({'V_m': parameter})

node_pos = np.array(nest.GetPosition(spatial_nodes))
node_pos[:,1]
v_m = spatial_nodes.get('V_m');

fig, ax = pyplot.subplots(figsize=(12, 6))
ax.plot(node_pos[:,0], v_m, '.', ms=3.5)
ax.set_xlabel('Node position on x-axis')
ax.set_ylabel('V_m');


NEST provides some functions to help create distributions based on for example the distance between two neurons.

#### Spatial distribution functions¶

The spatial_distributions module contains random distributions that take a spatial parameter as input and applies the distribution on the parameter. They are used for spatially distributed nodes.

Distribution function

Arguments

Function

nest.spatial_distributions.exponential()

x,
beta
$p(x) = e^{-\frac{x}{\beta}}$

nest.spatial_distributions.gaussian()

x,
mean,
std
$p(x) = e^{-\frac{(x-\text{mean})^2} {2\text{std}^2}}$

nest.spatial_distributions.gaussian2D()

x,
y,
mean_x,
mean_y,
std_x,
std_y,
rho
$p(x) = e^{-\frac{\frac{(x-\text{mean_x})^2} {\text{std_x}^2}+\frac{ (y-\text{mean_y})^2}{\text{std_y}^2}+2 \rho\frac{(x-\text{mean_x})(y-\text{mean_y})} {\text{std_x}\text{std_y}}} {2(1-\rho^2)}}$

nest.spatial_distributions.gamma()

x,
kappa
$p(x) = \frac{x^{\kappa-1}e^{-\frac{x} {\theta}}}{\theta^\kappa\Gamma(\kappa)}$

With these functions, you can recreate for example a Gaussian kernel as a parameter:

 NEST 2.x NEST 3.0 kernel = {"gaussian": {"p_center": 1.0, "sigma": 1.0}}  param = nest.spatial_distributions.gaussian( nest.spatial.distance, p_center=1.0, std_deviation=1.0) 
N = 21
middle_node = N//2

positions = nest.spatial.free([[x, 0.] for x in np.linspace(0, 1.0, N)])
spatial_nodes = nest.Create('iaf_psc_alpha', positions=positions)

parameter = nest.spatial_distributions.exponential(nest.spatial.distance, beta_Ca=0.15)

# Iterate connection to get statistical connection data
for _ in range(2000):
nest.Connect(spatial_nodes[middle_node], spatial_nodes,
conn_spec={'rule': 'pairwise_bernoulli',
'p': parameter})

targets = nest.GetConnections().get('target')

fig, ax = pyplot.subplots(figsize=(12, 6))
bars = ax.hist(targets, bins=N, edgecolor='black', linewidth=1.2)

pyplot.xticks(bars[1] + 0.5,np.arange(1, N+1))
ax.set_title('Connections from node with NodeID {}'.format(spatial_nodes[middle_node].get('global_id')))
ax.set_xlabel('Target NodeID')
ax.set_ylabel('Num. connections');


#### Mathematical functions¶

Parameter

Description

nest.random.exp(x)

Calculates the exponential of a parameter
nest.random.cos(x)

Calculates the cosine of a parameter
nest.random.sin(x)

Calculates the sine of a parameter

The mathematical functions take a parameter object as argument, and return a new parameter which applies the mathematical function on the parameter given as argument.

positions = nest.spatial.free([[x, 0.5*x] for x in np.linspace(0, 1.0, 100)])
spatial_nodes = nest.Create('iaf_psc_alpha', positions=positions)

parameter = -60 + nest.math.exp(nest.spatial.pos.x**4)
# Also available:
#   - nest.math.sin()
#   - nest.math.cos()

spatial_nodes.set({'V_m': parameter})

node_pos = np.array(nest.GetPosition(spatial_nodes))
node_pos[:,1]
v_m = spatial_nodes.get('V_m');

fig, ax = pyplot.subplots(figsize=(12, 6))
ax.plot(node_pos[:,0], v_m, '.', ms=6.5)
ax.set_xlabel('Node position on x-axis')
ax.set_ylabel('V_m');


#### Clipping, redraw, and conditionals¶

Parameter

Description

nest.math.min(x, value)


If a value from the Parameter is above a threshold, the value is replaced with the value of the threshold.

nest.math.max(x, value)


If a value from the parameter is beneath a threshold, the value is replaced with the value of the threshold.

nest.math.redraw(x, min, max)


If a value from the parameter is outside of the limits given, the value is redrawn. Throws an error if a suitable value is not found after a certain number of redraws.

nest.logic.conditional(x, val_true, val_false)


Given a condition, yields one value or another based on if the condition evaluates to true or false.

Note that x is a nest.Parameter.

The nest.math.min() and nest.math.max() functions are used to clip a parameter. Essentially they work like the standard min() and max() functions, nest.math.min() yielding the smallest of two values, and nest.math.max() yielding the largest of two values.

# This yields values between 0.0 and 0.5, where values from the
# distribution that are above 0.5 gets set to 0.5.
nest.math.min(nest.random.uniform(), 0.5)

# This yields values between 0.5 and 1.0, where values from the
# distribution that are below 0.5 gets set to 0.5.
nest.math.max(nest.random.uniform(), 0.5)

# This yields values between 0.2 and 0.7, where values from the
# distribution that are smaller than 0.2 or larger than 0.7 gets
# redrawn from the distribution.
nest.math.redraw(nest.random.uniform(), min=0.2, max=0.7)


The nest.logic.conditional() function works like an if/else statement. Three arguments are required:

• The first argument is a condition.

• The second argument is the resulting value or parameter evalued if the condition evaluates to true.

• The third argument is the resulting value or parameter evalued if the condition evaluates to false.

# A Heaviside step function with uniformly distributed input values.
nest.logic.conditional(nest.random.uniform(min=-1., max=1.) < 0., 0., 1.)

positions = nest.spatial.free([[x, 0.5*x] for x in np.linspace(0, 1.0, 50)])
spatial_nodes = nest.Create('iaf_psc_alpha', positions=positions)

spatial_nodes.set(V_m=nest.logic.conditional(nest.spatial.pos.x < 0.5,
-55 + 10*nest.spatial.pos.x,
-55))

node_pos = np.array(nest.GetPosition(spatial_nodes))
node_pos[:,1]
v_m = spatial_nodes.get('V_m');

fig, ax = pyplot.subplots(figsize=(12, 6))
ax.plot(node_pos[:,0], v_m, 'o')
ax.set_xlabel('Node position on x-axis')
ax.set_ylabel('V_m');


#### Combine parameters¶

NEST parameters support the basic arithmetic operations. Two parameters can be added together, subtracted, multiplied with each other, or one can be divided by the other. They also support being raised to the power of a number, but they can only be raised to the power of an integer or a floating point number. Parameters can therefore be combined in almost any way. In fact the distribution functions in nest.spatial_distributions are just arithmetic expressions defined in Python.

Some examples:

# A uniform distribution yielding values in the range (-44., -64.).
p = -54. + nest.random.uniform(min=-10., max=10)

# Two random distributions combined, with shifted center.
p = 1.0 + 2 * nest.random.exponential() * nest.random.normal()

# The node position on the x-axis, combined with a noisy y-axis component.
p = nest.spatial.pos.x + (0.4 * nest.spatial.pos.y * nest.random.normal())

# The quadratic distance between two nodes, with a noisy distance component.
p = nest.spatial.distance**2 + 0.4 * nest.random.uniform() * nest.spatial.distance


#### Use parameters to set node properties¶

Using parameters makes it easy to set node properties

NEST 2.x

NEST 3.0

for gid in nrns:
v_m = numpy.random.uniform(-20., 20.)
nest.SetStatus([node_id], {'V_m': V_m})

nrns.V_m=nest.random.uniform(-20., 20)


## What’s changed?¶

### Model parameters and their functionalities¶

Consistently use term synapse_model throughout:

As all PyNEST functions that used to take the list returned by Create now use the NodeCollection returned by Create, there shouldn’t be too many changes on the PyNEST level. One important change though, is that we now use synapse_model throughout to reference the synapse model.

Most importantly, this will change your Connect call, where instead of passing the synapse model with the model key, you should now use the synapse_model key.

>>>  nrns = nest.Create('iaf_psc_alpha', 3)
>>>  nest.Connect(nrns, nrns, 'one_to_one', syn_spec={'synapse_model': 'stdp_synapse'})


Simillarly, GetDefaults used to return an entry called synapsemodel. It now returns and entry called synapse_model.

Use allow_offgrid_times throughout:

In the model spike_generator, the parameter allow_offgrid_spikes is renamed allow_offgrid_times for consistency with other models.

Use unit ms instead of number of simulation steps:

The structural_plasticity_update_interval now has the unit ms instead of number of simulation steps.

### Topology module¶

• All topology functions are now part of nest and not nest.topology

• You can use the Create and Connect functions for spatial networks, same as you would for non-spatial network

• All former topology functions that used to take a layer ID, now take a NodeCollection

• All former topology functions that used to return node/layer IDs now return a NodeCollection

Note

See the reference section Function related to spatially distributed nodes in our conversion guide for all changes made to functions

All of the functionality of Topology has been moved to the standard functions. In fact, there is no longer a Topology module in NEST. The functions for creating spatially arranged neuronal networks are now in the nest module.

#### Create spatially distributed nodes¶

Creating spatially distributed nodes is now handled by with the standard nest.Create() function. Arguments of node creation have also been changed to make creating populations with and without spatial information more unified. To create nodes with spatial positions, nest.Create() must be provided with the positions argument

spatial_nodes = nest.Create(model, positions=spatial_data)


where spatial_data can be one of the following

• nest.spatial.grid()

This creates nodes on a grid, with a prescribed number of rows and columns, and, if specified, an extent and center. It can be easier to think of the grid as being defined by number of elements in x-direction and y-direction instead of thinking of rows and columns. Some example grid spatial nodes specifications:

nest.spatial.grid(shape=[5, 4], extent=[2., 3.])  # 5x4 grid in a 2x3 square
nest.spatial.grid(shape=[4, 5], center=[1., 1.])  # 4x5 grid in the default 1x1 square, with shifted center
nest.spatial.grid(shape=[4, 5], edge_wrap=True)  # 4x5 grid with periodic boundary conditions
nest.spatial.grid(shape=[2, 3, 4])  # 3D 2x3x4 grid

• nest.spatial.free()

This creates nodes positioned freely in space. The first argument to nest.spatial.free() can either be a NEST parameter that generates the positions, or an explicit list of positions. Some example free spatial nodes specifications:

nest.spatial.free([[5., 1.], [4., 2.], [3., 3.]])  # Three nodes with explicit positions

nest.spatial.free(nest.random.lognormal(),  # Positions generated from a lognormal distribution
num_dimensions=2)         # in 2D

nest.spatial.free(nest.random.uniform(),  # Positions generated from a uniform distribution
num_dimensions=3,       # in 3D
edge_wrap=True)         # with periodic boundary conditions


Note the following

• For positions generated from NEST parameters, the number of neurons has to be provided in nest.Create().

• The extent is calculated from the positions of the nodes, but can be set explicitly.

• If possible, NEST tries to deduce the number of dimensions. But if the positions are generated from NEST parameters, and there is no extent defined, the number of dimensions has to be provided.

spatial_nodes = nest.Create('iaf_psc_alpha', n=5,
positions=nest.spatial.free(nest.random.uniform(),
num_dimensions=3))


Spatially positioned nodes are no longer subnets, as subnets have been removed, but are rather NodeCollections with metadata. These NodeCollections behave as normal NodeCollections with two exceptions:

• They cannot be merged, as concatenating NodeCollections with metadata is not allowed.

• When setting the status of nodes and connecting spatially distributed NodeCollections you can use spatial information as parameters.

The second point means that we can use masks and position dependent parameters when connecting, and it is possible to set parameters of nodes based on their positions. We can for example set the membrane potential to a value based on the nodes’ position on the x-axis:

snodes = nest.Create('iaf_psc_alpha', 10
positions=nest.spatial.free(
nest.random.uniform(min=-10., max=10.), num_dimensions=2))
snodes.set('V_m', -60. + nest.spatial.pos.x)

Composite layers:

It is no longer possible to create composite layers, i.e. layers with multiple nodes in each position. To reproduce this, we now create multiple spatially distributed NodeCollections.

NEST 2.x

NEST 3.0

l = tp.CreateLayer(
{'rows': 1,
'columns': 2,
'elements':
['iaf_cond_alpha',
'poisson_generator']})

Use l when connecting, setting
parameters etc.

sn_iaf = nest.Create('iaf_psc_alpha'
positions=nest.spatial.grid(
shape=[2, 1]))

sn_poi = nest.Create('poisson_generator',
positions=nest.spatial.grid(
shape=[3, 1]))

Use sn_iaf and sn_poi when connecting,
setting parameters etc.


#### Retrieving spatial information¶

To retrieve the spatial information from your nodes, spatially structured NodeCollections have a .spatial parameter that will retrieve all spatial information as a dictionary.

>>>  spatial_nodes.spatial
{'center': (0.41717460937798023, 0.3541409997269511, 0.5058779059909284),
'edge_wrap': False,
'extent': (0.6786768797785043, 0.4196595948189497, 0.8852582329884171),
'network_size': 5,
'positions': ((0.1951471883803606, 0.24431120231747627, 0.5770208276808262),
(0.34431440755724907, 0.46397079713642597, 0.8201442817226052),
(0.17783616948872805, 0.4038907829672098, 0.16324878949671984),
(0.3796140942722559, 0.2643292499706149, 0.848507022485137),
(0.6565130492672324, 0.38219101540744305, 0.4020354822278023))}


Note that if you have specified your positions as a NEST parameter, NEST will convert that to a list with lists, and this is what you will get when calling .spatial.

#### Connect spatially distributed nodes¶

Similar to creating nodes with spatial distributions, connecting is now done with the standard nest.Connect() function. Connecting NodeCollections with spatial data is no different from connecting NodeCollections without metadata. In a layer-connection context, moving to the standard Connect() function brings with it some notable changes:

• Convergent and divergent specification of connection is removed, or rather renamed. See table below.

NEST 2.x

NEST 3.0

convergent

pairwise_bernoulli with use_on_source=True

convergent with num_connections

fixed_indegree

divergent

pairwise_bernoulli

divergent with num_connections

fixed_outdegree

use_on_source here refers to whether the mask and connection probability should be applied to the source neuron or the target neuron. This is only required for pairwise_bernoulli, as fixed_indegree and fixed_outdegree implicitly states if we are using the source or target nodes.

• The connection probability specification kernel is renamed to p to fit with pairwise_bernoulli, and is only possible for the connection rules in the table above.

• Using a mask is only possible with the connection rules in the table above.

#### Usage examples¶

A grid layer connected with Gaussian distance dependent connection probability and rectangular mask on the target layer:

NEST 2.x

NEST 3.0

l = tp.CreateLayer(
{'columns': nc, 'rows': nr,
'elements': 'iaf_psc_alpha',
'extent': [2., 2.]})

conn_dict = {'connection_type': 'divergent',
'kernel': {'gaussian':
{'p_center': 1.,
'sigma': 1.}},
{'lower_left': [-0.5, -0.5],
'upper_right': [0.5, 0.5]}}}
nest.ConnectLayers(l, l, conn_dict)

l = nest.Create('iaf_psc_alpha',
positions=nest.spatial.grid(
shape=[nc, nr],
extent=[2., 2.]))

conn_dict = {'rule': 'pairwise_bernoulli',
'p': nest.spatial_distributions.gaussian(
nest.spatial.distance,
std=1.),
{'lower_left': [-0.5, -0.5],
'upper_right': [0.5, 0.5]}}}
nest.Connect(l, l, conn_dict)


A free layer with uniformly distributed positions, connected with fixed number of outgoing connections, linear distance dependent connection probability and delay, and random weights from a normal distribution:

NEST 2.x

NEST 3.0

import numpy as np
pos = [[np.random.uniform(-1., 1.),
np.random.uniform(-1., 1.)] for j in range(1000)]
l = tp.CreateLayer({'positions': pos, 'extent': [2., 2.],
'elements': 'iaf_psc_alpha'})

conn_dict = {'connection_type': 'divergent',
'number_of_connections': 50,
'kernel': {'linear':
{'a': -0.5, 'c': 1.}},
'weights': {'normal':
{'mean': 0.0, 'sigma': 1.0}},
'delays': {'linear': {'a': 1.5, 'c': 0.}},
'allow_multapses': True,
'allow_autapses': False}
tp.ConnectLayers(l, l, conn_dict)

pos = nest.spatial.free(nest.random.uniform(-1., 1.),
num_dimensions=2)
l = nest.Create('iaf_psc_alpha', 1000, positions=pos)

conn_dict = {'rule': 'fixed_outdegree',
'outdegree': 50,
'p': 1. - 0.5*nest.spatial.distance,
'weight': nest.random.normal(mean=0., std=1.),
'delay': 1.5*nest.spatial.distance,
'multapses': True,
'autapses': False}
nest.Connect(l, l, conn_dict)


In NEST 3.0, the mask volume got removed, as the same mask was already available under the name box. The former was only an alias available in NEST for backward compatibility.

#### Retrieving distance information¶

If you have a SynapseCollection with connections from a spatially distributed network, you can retrieve the distance between the source-target pairs by calling .distance on the SynapseCollection.

s_nodes = nest.Create('iaf_psc_alpha', positions=nest.spatial.grid(shape=[3, 1]))
t_nodes = nest.Create('iaf_psc_alpha', positions=nest.spatial.grid(shape=[1, 3]))
nest.Connect(s_nodes, t_nodes)

conns = nest.GetConnections()
dist = conns.distance


.distance will be a tuple of the same length as your SynapseCollection, where dist[indx] will be the distance between the source-target pair at indx.

Calling .distance on a SynapseCollection where either the source or target, or both, are not spatially distributed also works, you will receive nan whenever one of the nodes is non-spatial.

### Recording from simulations¶

The spike_detector has been renamed to spike_recorder. The main rationale behind this is that the device is actually not detecting the occurence of spikes, but rather only records them. Moreover, the new name is more consistent with the naming of other similar devices that also end in the suffix _recorder.

In NEST 2.x, all recording modalities (i.e. screen, memory, and files) were handled by a single C++ class. Due to the many different responsibilities and the resulting complexity of this class, extending and maintaining it was rather burdensome.

With NEST 3.0 we replaced this single class by an extensible and modular infrastructure for handling recordings: each modality is now taken care of by a specific recording backend and each recorder can use one of them to handle its data.

NEST 3.0 supports the same recording backends for all modalities as in NEST 2.x. If compiled with support for SIONlib, an additional backend for writing binary files in parallel becomes available. This is especially useful on large clusters and supercomputers.

#### Changes¶

In NEST 2.x, the recording modality was selected by either providing a list of modalities to the record_to property, or by setting one or more of the flags to_file, to_memory, or to_screen to True.

In NEST 3.0, the individual flags are gone, and the record_to property now expects the name of the backend you want to use. Recording to multiple modalities from a single device is no longer possible. Individual devices have to be created and configured if this functionality is needed.

The following examples assume that the variable mm points to a multimeter instance, i.e., mm = nest.Create('multimeter') was executed.

 NEST 2.x NEST 3.0 nest.SetStatus(mm, {'record_to': ["file"]}) nest.SetStatus(mm, {'record_to': ["screen"]}) nest.SetStatus(mm, {'record_to': ["memory"]})  mm.record_to = "ascii" mm.record_to = "screen" mm.record_to = "memory"  nest.SetStatus(mm, {'to_file': True}) nest.SetStatus(mm, {'to_screen': True}) nest.SetStatus(mm, {'to_memory': True})  mm.record_to = "ascii" mm.record_to = "screen" mm.record_to = "memory"  nest.Create('spike_detector')  nest.Create('spike_recorder') 

You can retrieve the list of available backends using the following command:

list(nest.GetKernelStatus("recording_backends").keys())


Previously, the content and formatting of any output created by a recording device could be configured in a fine-grained fashion using flags like withgid, withtime, withweight, withport and so on. In many cases, this, however, lead to a confusing variety of possible interpretations of data columns for the resulting output.

As storage space is usually not a concern nowadays, the new infrastructure does not have this plethora of options, but rather always writes all available data. In addition, most backends now write the name of the recorded variable for each column as a descriptive meta-data header prior to writing any data.

The accumulator_mode of the multimeter has been dropped, as it was not used by anyone to the best of our knowledge and supporting it made the code more complex and prone to errors. In case of high user demand, the functionality will be re-added in form of a recording backend.

All details about the new infrastructure can be found in the guide on recording from simulations.

## What’s removed?¶

### Subnets¶

Subnets are gone. Instead NodeCollections should be used to organize neurons.

NEST 2.x

NEST 3.0

net = nest.LayoutNetwork(model, dim)
nrns = nest.GetLeaves(net)[0]

nrns = nest.Create(model, dim)


Printing the network as a tree of subnets is no longer possible. The PrintNetwork() function has been replaced with PrintNodes(), which prints ID ranges and model names of the nodes in the network.

NEST 2.x

NEST 3.0

>>>  nest.PrintNetwork(depth=2, subnet=None)
[0] root dim=[15]
[1]...[10] iaf_psc_alpha
[11]...[15] iaf_psc_exp

>>>  nest.PrintNodes()
1 .. 10 iaf_psc_alpha
11 .. 15 iaf_psc_exp


### Models¶

With NEST 3.0, some models have been removed. They all have alternative models that can be used instead.

Removed model

Replacement model

iaf_neuron

iaf_psc_alpha

aeif_cond_alpha_RK5

aeif_cond_alpha

iaf_psc_alpha_presc

iaf_psc_alpha_ps

iaf_psc_delta_canon

iaf_psc_delta_ps

subnet

no longer needed, use NodeCollection instead

Furthermore, the model iaf_tum_2000 has been renamed to iaf_psc_exp_htum. iaf_psc_exp_htum is the exact same model as iaf_tum_2000, it has just been renamed to match NEST’s naming conventions.

### Functions¶

Some functions have also been removed. The removed functions where either related to subnets, or they can be replaced by using other functions with indexing into a NodeCollection. The removed functions are (see also NEST 3.0: Detailed transition guide for a full list of functions that have changed):

• BeginSubnet

• ChangeSubnet

• CurrentSubnet

• DataConnect

• DisconnectOneToOne

• EndSubnet

• GetChildren

• GetElement

• GetLayer

• GetLeaves

• GetLID

• GetNetwork

• LayoutNetwork

• ResetNetwork

• RestoreNodes (have never existed on PyNEST level, it was just a SLI function)