Layers correspond to identifiable states in a cognitive model, often consisting of sets of homogeneous units.
some_layer description
The description of some_layer
will have the additional elements:
Representation
names the associated representation;
Width
indicates the width of that representation used to format the layer;
Current time
indicates the time currently associated with the layer;
Incoming connections
lists that attached connections with sub-elements indicating
the attached input channel as `port`;
Ports
list all input channels.
some_layer attach src_conn port
Establish src_conn
as an incoming connection on port port
(name) of some_layer
.
some_layer attach src_conn
As above, except the default port (net_input) is used.
some_layer detach src_conn
Detach src_conn
as connection to some_layer
on any port.
some_layer connections
List connections to some_layer
.
some_layer lesion
Insert a lesion_transfer
into the layer_transfer slot of some_layer
without destroying the existing transfer facet (which now becomes the
responsiblity of the lesion_transfer facet to manage; this is used
in the unlesion method).
some_layer unlesion
If some_layer
has a lesion_transfer
facet, replace this with the original
layer_transfer facet, and destroy the lesion_transfer
facet.
some_layer represent rep
Set the representaion of some_layer
to be rep
.
some_layer state time_step
Return the state of some_layer
at the (integer) time_step
index in the
array of states. Note time_step is the rank time, not the time (see
time_summary
).
some_layer state
Return the pattern_reference that (continually) refers to the current state pattern.
some_layer output time_step
Return the state of some_layer
at the (integer) time_step
index in the
array of outputs. Note time_step is the rank time, not the time (see
time_summary
).
some_layer output
Return the pattern_reference that (continually) refers to the current output pattern.
some_layer input time_step port
Return the input to the port
channel (number) of some_layer
at the
(integer) time_step
index in the array of inputs. Note time_step is the
rank time, not the time (see time_summary
).
some_layer output time_step
As above, but the default input (0; net_input) is used.
some_layer output
As above, but the current time is used.
some_layer input_ref port
Return the updating pattern_reference to the current input to some_layer
on port number port
.
some_conn set_lag nsteps
Set the amount of time between input and output of some_conn
to be
nsteps
. lag is the time differential between input and
state/output, such that the output at time t
can be generated with only
the input up to and including the time t-lag
.
some_conn get_labels
List the dimensions of the representation associated with some_conn
.
some_conn rep
Return a representation_reference to the representation of some_conn
.
some_conn time
Return a strength_pattern whose first and only element is the time at the layer.
some_conn time_summary
Print a human-readable description of the association between integer indices in the input/state/output arrays and the floating point times.
some_conn default_observer
Return a dflog_nobserver owned by the layer with default settings to observe
the
state of some_conn
.
some_conn default_input_observer port
Return a dflog_nobserver owned by the input_port with default settings to
observe
the
input channel port number port
of some_conn
.
some_conn default_input_observer
As above, but port number is 0 (net_input).