CM1K Hardware User’s Manual
20
Remark #1
: If the network is full, a learning operation will have no effect. You can detect that all the neurons of
the network are already committed by executing the Read NCOUNT command which will then return the value
0xFFFF.
Remark #2
: If an application requires to change learning settings such as the Minimum and Maximum Influence
Fields, this must be done prior to the broadcast of the vector to learn.
Remark #3
: If the AIF of a neuron reaches the Minimum Influence Field, the bit 15 of its category register is set to
1. The neuron is said “degenerated”. It still reacts to input patterns as any other committed neuron but the bit 15
of its category indicates that the neuron was prevented from shrinking its AIF to a smaller value during training and
its response should be weighted differently than the response of another firing neuron which is not degenerated.
Example
Let’s take the example of an input vector equal to a series going from 00 to 99. This vector has a length of 100
components. Its first 99 components are written in sequence to the CM_COMP register and its last and 100
th
component is written to the CM_LCOMP register.
Broadcast the vector, learn as category 33 and read the number of committed neurons.
Sequence
For (i = 0; i<99, i++) Write CM_COMP, Vector(i);
Write CM_LCOMP, Vector(99)
Write CM_CAT, 33
Read CM_NCOUNT, ncount
Ncount will not be incremented if a committed neuron with a category register equal to 33 already recognizes
Vector. Note that this does not mean that the content of the neurons has not changed. Indeed, no new neuron has
been committed but existing committed neurons may have reduced their influence fields.