Memory - Computational Explorations in Cognitive Neuroscience - page 298

play. If a particular cue stimulus is shown (e.g., an A),
then the subject has to remember the next stimulus, and
determine if it matches the one that comes two stimuli
after that. After every stimulus presentation, a button
must be pressed — one button if a match event has just
occurred, and another button otherwise. Thus, whether
one wants to encode a stimulus into active memory or
not depends dynamically on preceding stimuli, and can-
not be determined as a function of the specific stimu-
lus itself. Further, once encoded, the stimulus must be
maintained in the face of the two intervening stimuli.
Because of this need both to maintain robustly and
update rapidly, the working memory system cannot
adopt a consistent strategy for active maintenance —
it cannot always make the active memories robust by
making them insensitive to their inputs, because this
would preclude updating. Similarly, if the active mem-
ories are easily updatable as a function of their inputs,
they will not be robustly maintained in the face of irrel-
evant information on these inputs.
In this section we will see that the kind of simple
active memory system that we have been exploring is
missing the kind of dynamic switching between main-
tenance and updating that seems to be necessary. Thus,
the need for this kind of dynamic regulation system pro-
vides one more reason to believe that there is a spe-
cialized neural system for supporting activation-based
working memory. We will explore some ideas regard-
ing the nature of this specialized system and its dynamic
regulation in section 9.5.

same parameters cannot achieve both of these objec-
tives. Thus, we will explore how the parameters can be
manipulated to alter the network's tendency to maintain
or update.

To select this new environment, set env_type

, !

on the overall act_maint_ctrl control panel to

MAINT_UPDT instead of MAINT .
We also want to use the ISOLATED network as ex-
plored previously, because it provides the best active
maintenance performance.

Set net_type to ISOLATED (if it is not already).
To add realism and ensure that the basic maintenance
task is not completely trivial, let's also add noise.

Set the noise_var parameter to .01. View the
GRID_LOG (if not already open), and then Run the net-
work.
The first part is the same as before, but then the
Input2 input is presented, followed by the Maint2
maintenance period. Note that the grid log will scroll,
such that at the end of the run, only this last set of events
is shown. You can use the VCR-style arrow buttons at
the top of the log to scroll back and forth to view the
entire sequence.
You should observe in this case that the network up-
dates its internal representation upon the Input2 in-
put pattern presentation. If the task context at this
point called for the active maintenance of this new in-
put (e.g., in the CPT-like task described previously,
Input1 would be the cue stimulus in this case), then
this would be desirable behavior. However, it is also
possible that Input2 could be a transient bit of infor-
mation that should not be maintained (e.g, one of the
two intervening stimuli in the CPT-like task). In this
latter case, the network's behavior would be inappro-
priate.
The obvious parameter to manipulate to determine
whether the network robustly maintains or rapidly up-
dates is the relative strength of the recurrent self-
maintenance connections compared to the input con-
nections. The wt_scale parameter in the control
panel lets us adjust this, by determining the relative
strength of the recurrent self-maintenance connections.

Exploring the Model

Continue to use project act_maint.proj.gz (or
open it). Press the Defaults button to reset to default
parameters if already open.
We are now going to explore an environment that
starts out by presenting an input pattern and then re-
moving that input (as before), and then a new input pat-
tern will be presented and then removed. Under some
circumstances, we can imagine that the network would
want to update the active memory representations to re-
flect the second input, but in other circumstances, this
input may be irrelevant and should be ignored. It should
be clear at the outset that the same network with the

, !

Try setting the wt_scale parameter to 2 instead of
the default of 1, and Run a couple of times.