Data Analysis From Scratch With Python: Step By Step Guide

15. Artificial Neural Networks
For us humans it’s very easy for us to recognize objects and digits. It’s also
effortless for us to know the meaning of a sentence or piece of text. However,
it’s an entirely different case with computers. What’s automatic and trivial for us
could be an enormous task for computers and algorithms.
In contrast, computers can perform long and complex mathematical calculations
while we humans are terrible at it. It’s interesting that the capabilities of humans
and computers are opposites or complementary.
But the natural next step is to imitate or even surpass human capabilities. It’s like
the goal is to replace humans at what they do best. In the near future we might
not be able to tell the difference whether whom we’re talking to is human or not.
An Idea of How the Brain Works
To accomplish this, one of the most popular and promising ways is through the
use of artificial neural networks. These are loosely inspired by how our neurons
and brains work. The prevailing model about how our brains work is by neurons
receiving, processing, and sending signals (may connect with other neurons,
receive input from senses, or give an output). Although it’s not a 100% accurate
understanding about the brain and neurons, this model is useful enough for many
applications.
This is the case in artificial neural networks wherein there are neurons (placed in
one or few layers usually) receiving and sending signals. Here’s a basic
illustration 
from 
TensorFlow 
Playground
: 

---

Notice that it started with the features (the inputs) and then they’re connected
with 2 “hidden layers” of neurons. Finally there’s an output wherein the data was
already processed iteratively to create a useful model or generalization.
In many cases how artificial neural networks (ANNs) are used is very similar to
how Supervised Learning works. In ANNs, we often take a large number of
training examples and then develop a system which allows for learning from
those said examples. During learning, our ANN automatically infers rules for
recognizing an image, text, audio or any other kind of data.
As you might have already realized, the accuracy of recognition heavily depend
on the quality and quantity of our data. After all, it’s Garbage In Garbage Out.
Artificial neural networks learn from what feed in to it. We might still improve
the accuracy and performance through means other than improving the quality
and quantity of data (such as feature selection, changing the learning rate, and
regularization).