The inventors have always dreamed about creating a machine that
could think. These desires date back to at least the time of ancient
Greece. And when the programmable computing devices were first conceived,
people again started wondering if they could become so intelligent that
they might be able to develop their own thinking skills. Hundred years
down the road, and man built one such machine. Today, Artificial
Intelligence (AI) is a thriving field with many practical applications and
active research topics.
In the early days of
arti?cial intelligence, the ?eld rapidly tackled and solved numerous
problems that are intellectually di?cult for human beings. But, the
true challenge for AI proved to be the tasks that are easy for people to
perform, problems that we solve intuitively, that feel automatic to us,
like recognizing spoken words or faces in images.
What could be the solution to
such problems? It is to allow the computers to learn from their own
experiences. If they start getting knowledge from experience, the need for
human operators to specify all the required knowledge that the
computer needs will vanish. In other solutions, the computer also tries to
understand the world in terms of a hierarchy of concepts, with each
concept defined through its relation to simpler concepts. This enables the
computer to learn complicated concepts by building them out of simpler
ones. If a graph is drawn showing how these two concepts of learning
through experience and simplification of concepts are built on top of each
other, the graph is deep, with many layers. Therefore, this approach to AI
is called Deep Learning.
Broadly, AI is the
computer-based exploration of methods for solving challenging tasks that have
traditionally depended on people for solution. Such tasks include complex
logical inference, diagnosis, visual recognition, comprehension of natural
language, game playing, explanation, and planning.
Based Machine Learning
Machine MacHack Medical Diagnosis
Program Deep Learning
Commercial Expert Mobile Recommendation Intelligence System System Applications
Robot Behaviour Based Recommendation
Solver Robotics Technology
of Artificial Intelligence
Motivation Behind Deep Learning:
Deep Learning is a subset
of subset of Artificial Intelligence (Machine Learning). Ironically, abstract and formal tasks that are among
the most di?cult mental undertakings for a human being are among the
easiest for a computer. Computers have been able to defeat the best human
chess players for a long time but only recently, they have begun matching some
of the abilities of average human beings to recognize objects or speech. A
person’s everyday life requires an immense amount of knowledge about the
world. Much of this knowledge is subjective and intuitive, and therefore
di?cult to articulate in a formal way. Computers will also need to capture
this same knowledge in order to behave in an intelligent way. One of the
key challenges in arti?cial intelligence is how to get this informal
knowledge into a computer.
Most of the arti?cial
intelligence projects have sought to hard-code knowledge about the world
in formal languages. A computer can reason automatically about statements
in these formal languages using logical inference rules. This is known as
the knowledge based approach to arti?cial intelligence. But there have been many
difficulties in these systems relying on hard-code knowledge. We
have always struggled to devise formal rules with enough complexity to
accurately describe the world. Therefore, it was suggested that AI systems
need to acquire their own knowledge, by extracting patterns from raw data.
This capability is known as Machine
Learning. The introduction of machine learning enabled computers to
tackle problems involving knowledge of the world and make decisions that
The performance of the
simple machine learning algorithms depends heavily on the representation
of the data they are given. And this dependence is a general phenomenon
that appears throughout computer science and even daily life. In computer
science, operations such as searching a collection of data can
proceed exponentially faster if the collection is structured and indexed
intelligently. People can easily perform arithmetic on Arabic numerals but
?nd arithmetic on Roman numerals much more time consuming. It is not
surprising that the choice of representation has an enormous e?ect on the
performance of machine learning algorithms.
Learning is also, a specific kind of machine learning. It is
inspired by the structure of the human brain and is particularly effective
in feature detection. This also involves feeding the system with large
volumes of data. But since it is inspired by our brain, we first need to
understand how the neural network works.
Neural Networks and Working of Deep Learning:
A Neural Network passes data through interconnected layers of
nodes, like the vast network of neurons in the brain, classifying
information and characteristics of a layer before passing the results on
to other nodes in subsequent layers. The difference between a neural
network and a deep learning network is contingent on the number of layers: A
basic neural network may have two to three layers, on the other hand
a deep learning network may have dozens or hundreds of layers.
learning achieves great power and flexibility by learning to represent
the world as nested hierarchy of concepts, with each concept defined in
relation to simpler concepts, and more abstract representations computed
in terms of less abstract ones.
The most effective
results of Deep Learning can be observed in feature detection, and it does
so in the same way our brain intuitively detects different features. When
we try to recognize a square from other shapes our brain first
checks whether there are four lines associated with a figure or not. If it
finds four lines, it further checks if they are connected, closed,
perpendicular and that they are equal as well (Nested Hierarchy of
Concept). So, we take a complex task and brake it in simple, less abstract
tasks. This is exactly what Deep Learning does, but at a
Similarly, if we make a
system that must recognize whether the given image is of a cat or a
dog. If we try to solve this as a typical machine learning problem, we
will define features such as if the animal has whiskers or not, if the
animal has ears & if yes, then if they are pointed. In short, we will
define the facial features and let the system identify which features are
more important in classifying a particular animal.
Now, Deep Learning takes
this one step ahead. It can automatically find out the features that are
important for classification between a cat and a dog, whereas in Machine
Learning we had to manually give the features.
· Deep Learning
first identifies what are the edges that are most relevant to find out a Cat
or a Dog
· Further, it builds
on a hierarchically structure to find out what are the combination of all the
shapes and edges. For example, whether whiskers are present, or
whether ears are present, etc.
· After consecutive
hierarchical identification of complex concepts, it then decides which of
these features are responsible for finding the answer.
Over the last few years, Deep Learning has been
applied to numerous problems, ranging from computer vision to natural language
processing. In many cases Deep Learning outperformed previous work. Three
major reasons behind the breakthrough of (deep) neural networks are:
The availability of huge amounts of
Powerful computational infrastructure.
Advances in academia.
Since then, deep learning systems have started
outperforming not only the classical methods, but also some of the human
benchmarks in various tasks like image classification or face recognition.
This has created the potential for many new applications leveraging deep
learning to solve real-world problems, like:
Medical Image Analysis
Prediction and Diagnosis
Anomaly Detection and Security
Game Playing and many more.