Long-Term Memory Neural Circuits, Fast and Precise - Brain Theory from a Circuits and Systems Perspective - page 137

Fig. 8.9 Memorization
circuit for an attribute
in a word
Circuit Model for Memorizing New Memories
The neurons that convey cues for memory search may also serve as conveyers for
attributes to long-term memory. However, now they are not just a sampling of cues
but the entire image. Figure 8.9 shows how the cellular recall model can work for
memorization.
The circle with the Q represents a blank memory element. Memorization begins
when Memorization Enable goes true. However, for memorization to occur, a
control line termed the Memorization Control Line must be true.
Multiwrite Circuits
There have to be controls to select only one blank word so that not all available
words are programmed with the same information. This may be accomplished using
a neural multiwrite which is a neural form of a one-hot circuit. Memorization can be
regulated by signals to a Memorization Control Line, whose purpose is to give
multiwrite capability.
Multiwrite Using Long-Term Memory Elements
Figure 8.10 illustrates a simple system based on memory elements, the elementary
kind whose outputs are initially false, but which once set, the outputs stay true
indefinitely. The XOR at the bottom of the figure has one input held to true, so its
output to the first Memorization Control Line is true. All other memory elements for
control are resting at false, so they are removed from the system.