Daperture

Artificial Neural Networks (ANN) are fascinating. Really, they are just an attempt to focus on solving real-world classification problems by emulating what is known about the human brain. As complex as computers are, ultimately they work on a very simple concept. Pushing 0s and 1s around in the von neuman architecture. Simply put, they process very specific commands one by one, sequentially as presented by their programmer. The logic in how our complex computer works is fairly simple, if this is 1, do this otherwise do something else.

When it comes to problems of classification, things can get much more complex. How would you write a program to classify a given text's language? Is it English, French or German? One could look for specific words respective of each language, but what about an English work that references a German word? Ultimately, a programmer would need to look for patterns in each language and search for the patterns to classify a piece of text. This could prove easy, but it's far too easy to overlook more subtle patterns and discrepancies. Also if the patterns only vary by very slight amounts, it can get very complicated to build more and more logic and checks into a classification program. ANN aids to solve such problems by suggesting that you could feed varying language texts into an ANN with the correct result. This training would to the ANN would allow it to automatically figure out different patterns in each language and correctly classify text it has never seen before.

It's almost like how our mind works. How does one realize that an individual is a certain race just by seeing their face? There's absolutely no way to know all 6 billion faces that exist and have our minds figure out that they are seeing face #139392425 who is Irish. We intuitively pull specific characteristics to and see that someone meets a set of criteria specific to a given race and most times are correct. Our life experiences serve as our training data, 'Oh, this is what an Asian looks like, while this is what a mixed caucasion/asian looks like.'

On a very high level, a NN works by being made up of many neurons that are all connected to each other. Each neuron communicates with each other by sending electrical impulses from one to another, however they the neurons don't immediately act when an electrical impulse is received. Only when certain thresholds are crossed, does a neuron then maybe send out another electrical impulse to neighboring neurons. So, it takes a bit of experience for the neurons in our brains to register the action needed to be taken for certain events, however our minds adapt and they learn.

I think in the nature versus nurture argument, neural passage ways and "training data" so to speak are definitely passed down from generation to generation. Anecdotal evidence is seen in how certain observation are made, "Wow, Johnny has the same walk, same gestures as his Father. He even has such a similar personality!"

This would aid in the theory that we have a predisposition for certain tendencies such as alcoholism, sexual orientation, etc. However at the same time, I'd like to suggest that as predisposed as we are, our minds are powerful enough to be malleable to outside input and experiences. Entertaining that two individuals have an equal footing on will power, it seems that for certain genetic training data is far more ingrained in some people than others. One may easily adapt to enjoy certain distastes while the other may not be so easily inclined to adapt/change.

Despite natural predispositions passed on by our parents, ultimately we are individuals responsible our own choices and actions. How do we cope with personal predispositions that aren't socially acceptable? Also, how do we cope with other's socially unacceptable predispositions? I'll throw it out there. Is it okay to be homo-sexual?