void make_layer_recurrent(Layer* layer)
{
vector<int> delay = list_of<int>().repeat(layer->num_seq_dims(), 0);
loop (int i, indices(delay))
{
delay[i] = -layer->directions[i];
connect_layers(layer, layer, delay);
delay[i] = 0;
}
/*
BeginDocumentation
Name: topology::ConnectLayers - connect two layers
Synopsis: sourcelayergid targetlayergid connection_dict
ConnectLayers -> -
Description: Connects nodes in two topological layers.
The parameters set in the input dictionary decides the nature
of the connection pattern being created. Please see parameter
list below for a detailed description of these variables.
The connections are created by iterating through either the
source or the target layer, consecutively connecting each node
to a region in the opposing layer.
Parameters:
sourcelayergid - GID of source layer
targetlayergid - GID of target layer
connection_dict - dictionary containing any of the following
elements:
------------------------------------------------------------------
Connection dictionary parameters:
------------------------------------------------------------------
Parameter name: connection-type
Type: string
Parameter description:
Decides the type of connection pattern being created (i.e.
convergent or divergent topological connection). A convergent
topological connection is a connection between a source region
and a target node. A divergent topological connection is a
connection between a source node and a target region. A convergent
topological connection can also be called a receptive field connection.
A divergent topological connection can also be called a projective
field connection. A one-to-one connection can be created by setting
the size of the source or target region equal to one. The connection
type has particular effect on the connection pattern when used together
with the number_of_connections variable.
Parameter name: mask
Type: dictionary
Parameter description:
The mask defines the region used in the connection type described
above. There exists a selection of many different region sizes and
shapes. Examples are the grid region, the rectangular, circular or
doughnut region.
The grid region takes an optional anchor parameter. The anchor
parameter indicates which node of the grid region is aligned with
the source node.
Parameter name: weights, delays and kernel
Type: dictionary
Parameter description:
These parameters can be initialised in many ways. Either as a constant
value, with the help of a dictionary, or in an array (only for fixed
grid layers). The dictionary can be of type gaussian, 2D gaussian,
linear, exponential and other.
Parameter name: source
Type: dictionary
Parameter description:
The source dictionary enables us to give further detail on
how the nodes in the source layer used in the connection function
should be processed.
Parameters:
model* literal
lid^ integer
*modeltype (i.e. /iaf_neuron) of nodes that should be connected to
in the layer. All nodes are used if this variable isn't set.
^Nesting depth of nodes that should be connected to. All layers are used
if this variable isn't set.
Parameter name: target
Type: dictionary
Parameter description:
See description for source dictionary.
Parameter name: number_of_connections
Type: integer
Parameter description:
Maximum number of connections that each iterating node is allowed.
The actual connections being created are picked at random from all
the candidate connections.
Parameter name: synapse_model
Type: literal
Parameter description:
The synapse model to be used for creating the connection.
.
Parameter name: allow_autapses
Type: bool
Parameter description: Used together with the number_of_connections option to
indicate if autapses are allowed.
Parameter name: allow_multapses
Type: bool
Parameter description: Used together with the number_of_connections option to
indicate if multapses are allowed.
------------------------------------------------------------------
Example:
topology using
%Create source layer with CreateLayer
<< /rows 15
/columns 43
/extent [1.0 2.0]
/elements /iaf_neuron
>> /src_dictionary Set
src_dictionary CreateLayer /src Set
%Create target layer with CreateLayer
%%Create layer
<< /rows 34
/columns 71
/extent [3.0 1.0]
/elements {/iaf_neuron Create ; /iaf_psc_alpha Create ;}
>> /tgt_dictionary Set
tgt_dictionary CreateLayer /tgt Set
<< /connection_type (convergent)
/mask << /grid << /rows 2 /columns 3 >>
/anchor << /row 4 /column 2 >> >>
/weights 2.3
/delays [2.3 1.2 3.2 1.3 2.3 1.2]
/kernel << /gaussian << /sigma 1.2 /p_center 1.41 >> >>
/sources << /model /iaf_neuron
/lid 1 >>
/targets << /model /iaf_neuron
/lid 2 >>
/synapse_model /stdp_synapse
>> /parameters Set
src tgt parameters ConnectLayers
Author: Håkon Enger, Kittel Austvoll
SeeAlso: topology::CreateLayer
*/
void
TopologyModule::ConnectLayers_i_i_DFunction::execute( SLIInterpreter* i ) const
{
i->assert_stack_load( 3 );
index source_gid = getValue< long_t >( i->OStack.pick( 2 ) );
index target_gid = getValue< long_t >( i->OStack.pick( 1 ) );
const DictionaryDatum connection_dict = getValue< DictionaryDatum >( i->OStack.pick( 0 ) );
connect_layers( source_gid, target_gid, connection_dict );
i->OStack.pop( 3 );
i->EStack.pop();
}
