CM1K Hardware User’s Manual
27
3.5.7
Loading the contents of the neurons
Loading a knowledge to the neurons is made in the following order:
-
The first operation consists of clearing the existing neurons, setting the CM1K chip in Save_and_Restore
mode and pointing to the first neuron of the chain.
-
For each neuron, you can write its components, context, minimum influence field and active influence
field in any order. The category register must be written last since the instruction automatically points to
the next neuron in the chain.
-
Finally, when the neurons have been loaded, the last operation consists of setting the CM1K chip back to
its normal operation mode.
Remark 1
: Note that in Save_and_Restore mode the
last component is written to the CM_COMP register and
not to the CM_LCOMP register.
Remark 2
: If it is known that the patterns have a length
M with M<256, the number of Write COMP can be
limited to M, thus speeding the restore operation.
Remark 3
: If you intend to use the newly committed
neurons as a KNN classifier (as opposed to the default
RBF classifier), writing the AIF register is not necessary
since it will not be used by the KNN classifier.
Remark 4
: Loading the MINIF is necessary if it is known
that additional training will be done at a later time to
complete or expand the knowledge.
Remark 5
: There are few cases where loading several
knowledge bases into a same chain of neurons can be
relevant. One example consists of merging neurons’
contents associated to different contexts and thus
trained independently. If you are very cautious and
clearly understand the consequences of appending the
content of neurons to a knowledge already residing in a
chip, you can discard the Write Forget command. In
such case the neuron pointed after the Write NSR will
be the first neuron available in the chain or the RTL
neuron.
Remark 6
: If the content loaded into the neuron includes context values other than the default of 1, it is necessary
to execute a Write GCR at the end of the sequence so the newly loaded neurons can determine if they belong to
the currently active context or not.