Diving into A.I. for the First Time

1. Artificial Intelligence

What is A.I. ?

General Artificial Intelligence is a term used to describe ‘intelligence’ demonstrated by machines. By ‘intelligence’ we usually mean that the machine is familiar with making “cognitive” functions that resembles at a certain point with the Human Mind, such as “Learning” and “Problem solving”.
Leading A.I. Books available out there, define the field as the study of “intelligent agents”: any machine that perceives its environment and takes those actions that results in maximizing its chance of getting successful at achieving its goals.

The field was founded on the assumption that “human intelligence can be so precisely described that a machine can be made to simulate it”.

It is the superset of the fields that recently have gained more popularity, like Machine learning and Deep learning.

Applications

As machines become increasingly capable, tasks considered to require “intelligence” are often removed from the definition of AI, a phenomenon known as the AI effect.
For example, optical character recognition (OCR) is frequently excluded from things considered to be AI, having become a routine technology.
Modern machine capabilities generally classified as AI include successfully understanding human speech, competing at the highest level in strategic game systems (such as chess and Go), self-driving cars, intelligent routing in content delivery networks, and military simulations.

Approaches include statistical methods, computational intelligence, and traditional symbolic AI. The AI field draws upon computer science, information engineering, mathematics, psychology, linguistics, philosophy, and many other fields.

2. Machine Learning

What is Machine Learning ?

Machine learning is a field related to the study of computer algorithms that improves itself through experience with time (Also known as ‘Training’). Unlike typical computer algorithms, machine learning algorithms follows the concept of learning Rules based on past data and its already collected result. Let me explain this deftly.

Explaining the Diagram above:
The first part shows that if we input data or Problem scenario with rules (Also known as Step by Step method to calculate resultant output) into a typical algorithm engine than we get resultant values out of them. It’s pretty common in terms of applications and usage in daily world.
But with respect to Machine learning, this concept involves a structural change in the process cycle. In place of Rules, we input results that were already obtained and known to us through calculations or observations performed in the past with some input data. Applying these structural changes and modifying our typical algorithms, we can obtain the rules that are applied on the input data to generate such specific results from it. Those rules than can further be used on future data or un-calculated data to infer the results.

Machine Learning in classified into 3 Types:

Supervised Learning
Unsupervised learning
Reinforcement learning

Their usage depends solely on problem statements that you are dealing with, still most common form of machine learning used in the industry today is “supervised Learning”.

These various types are itself a topic of individual content and discussion, so we will get a glimpse about them in the upcoming contents.

Applications:

Regression
Classification
Clustering
Active learning
simulation based optimizations
single / Multi agent systems
and Many more

3. Deep Learning

What is Deep Learning ?

Personally speaking, this is the field of great Passion for me.

Deep learning is a subset field of Machine learning, dealing with Neural Networks. To understand the field, we need to first look into What is a Neuron and What is a Neural Network.

Neuron is the simplest and smallest structural unit in a Neural Network. It is nothing more than a couple of mathematical equations that creates its significant value as a part of any neural network. Here is a single Neuron:

These are 2 Equations involved in constituting a single Neuron. Let’s say what they are:

Wx + B = Z

: The First part of this equation is a matrix multiplication between the input values and weight matrix (every input value has a corresponding weight value for each neuron it is connected to. Note: Here in a single neuron, the $W$ and $x$ are single values due to the presence of a single neuron.
This weight value is like the regulator whose optimum value results in the best output through the model ). The second part of the equation is the summation of a constant term ( BIAS ) with the result of first part to shift the equation to the best position for it to fit the Input values.

$f(Z) = Y$ : The first equation, discussed above is used to identify the linear nature in the input data, as It is a simple linear equation. But If the input data is non-linear in nature, the ‘ $f()$ ’ or activation function is used to determine that non-linearity and fit the input data accordingly to the output equation.

Neural network is the combination of such Neurons with all sort of structure, varying in number, design, activation functions and many more. There are tons of various neural network structures created and still being researched on actively. Different designs are used for different use cases (for eg. working with Images, Text, Sequences etc.)
Here is a basic multilayer neuron network, consisting of

an Input layer,
a Hidden layer( the layer that is working in-between the input and output layer and whose values are hidden to user and only used to find the best results that are eventually useful to the user),
and an output layer.

Adding more and more of these hidden layers and making this network deep, is what gives the field its name: Deep Learning.

Applications

Image recognition
Time series forecasting
Recommendation systems
Image generation
Language translation
and many more really mesmerizing applications.

If you read the post till the end, I am grateful to you for showing interest and support. I will be coming with more content on topics related to Artificial Intelligence and Data Science. Please provide your valuable feedback to improve our content quality and reach more people.

Author: Gargeya Sharma

3rd year B.Tech Computer Science Student
Field of Study: Deep Learning, Data Science
Github
LinkedIn

Icons made by Flat Icons from www.flaticon.com

Search This Blog

Data Science & Deep Learning