Neural Networks 101
This is an artist's interpretation of two neurons communicating.
Note: These parts are not drawn to scale. In the body, the axon
can be one meter or longer.
Assuming you aren’t a current psychology or medical
student, you may be a little fuzzy on the way our body transmits
messages through the nervous system. So, here is a crash course
on how the process works — just in case high school biology was
a little further back in the past than you care to admit:
• Neurons are the cells in your body that help the brain
communicate with the muscles, and vice versa. The cells
communicate with each other using chemicals (called
neurotransmitters) that are transmitted from neuron to neuron.
• Neural networks work by sending chemicals across the gap
between the end bulbs of one neuron and the dendrites of the
other, called a synapse. Once the message has been sent across
the synapse, the journey ends for that batch of neurotransmitters.
After the message has been communicated down the next
neuron, that neuron makes more chemicals for the next neuron,
and so on.
Picture a line of dominoes. The first domino causes the next
one to fall, then the next, all down the line. This system differs
from an electrical one, where the same signal is transmitted all
the way down the line. It is crucial that the signal is “recreated” at
each step of the process. Otherwise, a signal sent from a neuron
in your shoulder to the brain would be stronger than one sent
from your toes. In other words, your shoulder would be more
sensitive to touch than your toes, which is not how we want our
bodies to work!
• Neurons require a certain amount of stimulation in order
to fire to pass their threshold. Say, a neuron needs a signal of at
least + 2 to fire (this is completely arbitrary). If it only gets a signal
of +1, the message won’t be sent to the next neuron. However,
the effects of neurons can accumulate. If a neuron is getting a
signal of +1 from three different neurons, then the effects will
add up and the neuron will send the message. Similarly, if a single
neuron keeps sending a signal of +1 over and over again, the
signals will add up to go past the threshold.
Here’s the thing: Sometimes neurons send excitatory signals
(ones that tell the neuron to fire) and sometimes they send
inhibitory signals (ones that say not to fire). These inhibitory and
excitatory signals can add up in countless ways, thus creating a
huge variety in the way our body communicates with itself.
One of the more esoteric ways human anatomy has
inspired artificial intelligence work is the Artificial Neural
Network (ANN). This is not a neural network in the literal
sense, so don’t go thinking that scientists are developing
robots that can feel, smell, and taste (although a robot has
been developed in Japan that can simulate feeling pain in
order to train dental students to be more gentle when
dealing with patients, but this isn’t the same thing).
A model of a simple artificial neural network.
These days, designers link together thousands
of neurons in millions of ways to create
50 SERVO 08.2013