The Missing Keys

By Antonio Gino, Ph.D

Recent statistics indicate up to 70 percent of Americans are exposed to some form of trauma during their life time. Most victims of trauma experience at least some symptoms of posttraumatic stress (PTS): intrusive thoughts, flashbacks, anxiety, vigilance, disturbing dreams, avoidance of reminders, survival guilt, anger issues, self-medication (usually with alcohol or other substances), sexual issues, etc. About 20 percent of trauma victims develop posttraumatic stress disorder (PTSD), i.e., e

• Language: English
• Publisher: Fulton Books, Inc.
• Release date: Apr 26, 2017
• ISBN: 9781633384002

Book preview

Neural Networks

The evolution of the human brain, like the evolution of all living things, including their organs, is related to the two fundamental needs so well elucidated by Charles Darwin: survival and reproduction. The brain is a highly specialized yet versatile organ of our body and has evolved powerful pattern recognition abilities to detect and process signals of danger, sources of food, and opportunities to reproduce. Through thousands of generations, via trial and error, our brain has evolved the ability to differentiate beneficial from nefarious foods, likely partners for reproduction from unlikely ones, areas where we can feel safe from dangerous ones, and animals or individuals we can trust from ones we must avoid.

The evolutionary advantage animals have gained via their ability to recognize patterns cannot be underestimated. The first primate who recognized a fig tree from similar-looking trees, which did not produce fruit, was much more likely to survive and leave offspring. The first lions to recognize a sick water buffalo were more likely to survive and leave offspring due to the reduced risk associated with attacking weakened prey. On the other hand, our ancestors who failed to recognize potential sources of danger were more likely to be killed before they had a chance to reproduce and leave behind viable offspring, i.e., their genetic heritage went extinct.

The question of just how the brain manages to detect a pattern is very complicated. Many books have been written on the subject and I will not reiterate what many authors have presented so well. However, the concept of a neural network is fundamental to understanding just what I mean by assumptions, so I will try and describe this concept.

Pattern Recognition

The human brain contains at least ten billion nerve cells or neurons. Each neuron has a nucleus and some or many dendrites. Dendrites are appendices, which reach out from the individual neurons and connect them to other neurons via electrochemical signals that travel through them. Individual neurons have an incredible variety with regard to number of dendrites: some neurons have only a few dendrites and some have more than sixty thousand. Obviously, some neurons need to connect to more neurons than others, i.e. all neurons are not created equal! The figure below, which is in the public domain, gives an illustration of how neurons are connected in the brain. Note how some neurons have few dendrites, while others have many dendrites (or potential connections).

Neural_Network_Cajal

From: Texture of the Nervous System of Man and the Vertebrates by Santiago Ramón y Cajal. The figure illustrates the diversity of neuronal morphologies in the auditory cortex. Found at http://www.anat.ucl.ac.uk/research/linden/

Different parts of the brain are also typically found to be specialized to process some aspects of signals we receive. For example, the back of the brain specializes in the processing of visual signals (visual cortex), the right hemisphere of the brain is usually more adept at processing spatial information, the left hemisphere focuses on speech, etc. Which parts of the brain are involved in what aspects of information processing is not directly relevant to our current topic. I mention the various specializations to emphasize many parts of the brain may be involved in the processing of signals.

An extremely simplistic way to look at how neural networks may lead to pattern recognition is to think of them as groups of neurons, which detect signals and fire other neurons, which recognize the pattern. For example, we could imagine three neurons firing to detect the letter A: one to detect the left rising line (/), one to detect the right falling line (\), and one to detect the middle bar (-). When these three neurons fire simultaneously, they generate electrochemical signals, which travel through the dendrites and stimulate the part of the brain that recognizes the spatial arrangement of the three lines and the part of the brain we have trained to respond with A. Again, this is an extremely simplified way of looking at how neural networks lead to pattern recognition. Clearly, even the recognition of the letter A requires the interaction of different areas of the brain and millions of neurons. For example, before we can recognize the letter A, the brain must have stored all information required to process the Roman alphabet to differentiate between upper and lower case letters, to ignore all information currently irrelevant to the letter A, to differentiate the letter A from similar looking letters, to associate the verbalization of the letter with its visual characteristics, etc.

Some patterns are much more complex than the previous example. For example, my younger sister recently asked me to look at a picture of a group of individuals from fifty plus years ago, and I was able to recognize four individuals from my youth, including myself after confusing myself with my older brother (we used to look very much alike when under seven years of age). Even a cursory explanation of how I was able to recognize the four individuals makes clear the complexity of the neural networks involved. I had to somehow recall how the four individuals used to look fifty plus years ago, I had to differentiate the looks of each individual from the others, I had to recall the historical context of the picture, etc. The number of neurons involved in such pattern recognition is in the many millions, possibly billions. Nevertheless, whether simple or complex, pattern recognition involves the same process: when neurons fire in a certain pattern, we have the ah! phenomenon, i.e. we recognize the pattern/object/situation.

Innate vs. Learned Pattern Recognition

The brain of most living organisms, including humans, seems to have the ability to recognize some patterns without any training and without even the awareness of recognizing the patterns. For example, recent research indicates we (humans) find attractive—i.e., recognize as potential mates for reproduction—members of the opposite sex who have immune systems different from our own. Mating between individuals with different immune systems increases the likelihood offspring from such mating will benefit from having immune systems which combine the two different ones of the parents, i.e., will result in offspring more likely to produce viable offspring themselves due to reduced susceptibility to disease. How we manage to recognize such patterns is a mystery that will require many years of research to be solved, and it is obviously not something we are taught.

Patterns we need to learn to recognize, on the other hand, require anything from minimal to very extensive training. For example, we seem to easily learn to recognize human faces or to avoid certain foods, e.g., it usually takes just one trial for us to learn a particular food made us ill and we avoid it on subsequent meals. On the other hand, we need many years of formal training to learn to recognize the words you are reading, areas of a city we need to avoid due to potential danger, or people we need to avoid.

Characteristics of Neural Networks

Neural networks have many characteristics and I leave it to network engineers to elucidate them. For the purposes of the current presentation, I will emphasize only two characteristics: growth and priming (or spreading activation).

The growth process of neural networks is easily illustrated by the way we learn a new language. Initially, it is a struggle to recognize and remember individual syllables and words. Once we master individual words, we struggle to remember sentences, grammar, and syntax. Once we master sentences/grammar/syntax, we can converse in ways we did not think was possible when we started our journey to learn the new language. We can visualize the interconnections (dendrites) between neurons to grow as we learn the language. At first, growth is limited to connections between individual letters, which form syllables. Next, the connections grow to words and their meaning (or translation). The growth then spreads to groups of words and their order. Finally, the growth spreads to sentences, grammar, and syntax.

We can imagine the above process to be analogous to the one we observe during the creation of an electrical grid (another word for network). At first, the electrical cable spreads from the power station to one house, then to another house, then to another house, then to another street, then to another neighborhood, then to an entire city. We can then connect the cables from the power station (same neuron) to other cities (new dendrites), then connect cities with cities, until we call the resulting grid a connected county. We finally connect the cables between power stations and counties and call it a connected state or nation. We can visualize neurons as spreading connections (dendrites) in the same fashion. The neurons involved in the recognition of a human face are connected to the neurons involved in the recognition of humans, which are connected to neurons involved in the recognition of apes, which are connected to neurons involved in the recognition of primates, which are connected to neurons involved in the recognition of mammals, etc. The network is both extensive and extensible. New connections between neurons are created as we learn more about a subject, like extending electricity to the new house in the neighborhood. If you did not know apes are higher order primates (or primates without a tail), your brain would create the connections (new dendrites) needed to associate the word/concept ape with the word/concept primate.

The growing network is a marvel of nature, although it has some disadvantages. As the network becomes more complex, it is also more likely to be subject to widespread problems. An electrical grid connecting a few houses can become impaired and affect only a few families. On the other hand, a cable system connecting entire cities can cause disruptions affecting entire cities, not to mention entire countries. We have seen this problem recently, when the Internet connections between countries have become disabled due to natural or man-made causes.

The second important characteristic I emphasize is the ability of networks to go into a primed or partially activated state. An example will again illustrate the concept. Experiments show we are faster at recognizing the word trunk after we see the word elephant than after we see the word lion. The explanation for the difference in reaction times is that seeing the word elephant activates neurons linked with the images/sounds/meaning/characteristics of elephants. Once associated neurons are at least partially activated, they are more easily reactivated when the word trunk appears, i.e., they remain partly active or primed after we see the word elephant.

Priming is an extremely important characteristic of networks for what I mean by assumptions. We are not conscious of the activation of related neurons when we see the word elephant. We do not expect or even suspect we may see the word trunk just because we saw the word elephant. Our brain activates related neurons automatically, without any need to involve thinking or any other process under our control, which brings us to the need for a brief introduction to controlled vs. automated brain functions.

Controlled vs. Automated Processes

The brain is constantly engaged in a myriad of functions. While you are reading this page, your brain is processing the meaning of the words you read, it is storing information regarding your surroundings, it is creating all connections necessary to retrieve (or reconstruct) the current information at a later date, and it is processing sounds, images, smells, and other stimuli in the environment. We are normally not aware of the active processing of all these stimuli unless a signal disrupts the process, e.g., we become immediately aware of our surrounding in case a smell indicates the possibility of fire.

We know the brain is processing all the apparently irrelevant information because we can later recall details we never thought we had stored. For example, we can recall we were reading while seated in our favorite chair, or while the radio was playing our favorite music. Analogously, we remember what we had for dinner last night and where we had dinner even though we never made any effort to store such information. Our brain performs all this automatic work in the background, without any