# Machine Learning – Introduction , Approaches & Applications

Machine learning and Artificial Intelligence have become the favorite topics of many scientists , engineers and even new comer computer students . Many argument and discussion are taking place regarding machine learning . The sole theory or definition of machine learning is still confusing . So , let’s talk a little bit about machine learning .

# What is Machine Learning ?

Machine learning ( sometime also termed as ML )  is a branch of computer science and sub-field of artificial intelligence which typically deals with machines that can learn themselves not by being explicitly programmed but analyzing huge amount of data .

The term Machine Learning ( ML ) was coined by Arthur Samuel in 1959 while at IBM . According to Samuel , ML is something that gives computer ability to learn without being explicitly programmed . A huge amount of data is provided and ML algorithm finds a specific pattern of data and finally analyzes new data of same type . The algorithm is made in such a way that it builds a model from sample input and later use it to make prediction on data . It is closely related to Computational Statistics and Mathematical Analysis as it works with huge data and analyses mathematically . ML relies on specific representation of data such like feature or characteristic that a computer can understand . The process of finding specific representation of data is also known as “feature extraction” . Note that the data sets provided as input must be similar to the data we are expecting ML to predict later .

## Types of Machine Learning (ML)

ML is classified according to the type of problems and tasks . It is classified into main three categories , depending on the nature of the learning “signal” or “feedback” available to a learning system.

1. Supervised Machine Learning
Supervised ML works with data in which specific/true label or feature or class is indicated . Let us take an example of ML in which we want to teach computer to distinguish dog and tiger . We can store a huge number of different pictures of dog and tiger in which their corresponding name i.e.”dog” and “tiger” is assigned as tag in each picture of them . Now we have true label in our data and we can use it to supervise our ML algorithm to distinguish the image of dogs and tigers . Once the algorithm can find the right image in our data , it can now tell us about the new image of either dog or tiger we want our ML program to tell us  . In this way , computer learns to predict about new data .
2. Unsupervised Machine Learning
Unsupervised ML is quite difficult than supervised as it deprive the learning algorithm of labels . Just huge data set is provided and algorithm is intended to find a pattern or lets say hidden characteristics itself . This method is used when labeling is not found in data . First data is separated in two categories labeled and not labeled and supervised learning is applied for labeled and unsupervised learning is applied for not labeled data . This process is called Clustering . For example , if we forgot to label “dog” or “tiger” on some images , then we apply unsupervised learning method to these . We make algorithm to distinguish dog’s and tiger’s image depending on their other characteristics such as their size , body structure , body color etc . The machine learning workflow for supervised and unsupervised learning can be explained by following picture .
3. Reinforcement Machine Learning
Reinforcement is another class of ML problems . It is completely different from supervised and unsupervised learning . It seeks for a result data which has already been received after performing certain task . Lets explain with an example . Suppose our ML is intended to play chess and learn itself to win . At first it has nothing to examine but after playing few games , it takes a move analyzing previous result of game . It stores the result of whole game in each completion of game and takes the best move . The more it plays chess , there is more chance of winning the game .

# Approaches Of Machine Learning

Some major approaches of machine learning are :-

• Deep Learning
• Artificial Neural Network
• Clustering
• Representation Learning
• Reinforcement Learning
• Genetic Algorithm
• Metric Learning
• Rule Based Learning
• Learning Classifier Systems
• Decision Tree Learning
• Association Rule Learning

# Applications Of Machine Learning

Machine Learning has great application to modern innovating and techno-science era . The list can not fit all applications . Some major applications of machine learning are :-

• Natural Language Processing
ML acts as main component of Artificial Intelligence (AI)  on Natural Language Processing . Machine Translation , Text Simplification , Question Answering , Speech Synthesis etc are expected to be achieved by means of ML .
• Recognition Technology
ML can make AI very simpler by achieving the goals : Voice Recognition , Hand Writing Recognition  , Pattern Recognition Image Recognition Facial Recognition Optical Character Recognition …etc .
• Search Engine
ML has huge application on Search Engine as it deals with large number of data .
• Data Mining
Data Mining is the hot topic on ML system . It is basically the process of finding special pattern on data and splitting similar data sets together .
• Email Filtering
ML helps find scam emails or unnecessary emails and puts in other place like spam folder in gmail .
• Other Fields
As I already mentioned , ML has applications in large scale and this single article can not list all these . I am gonna put just some topics . Some topics include Computer Vision , Customer Relationship Management , Inverted Pendulum , Recommendation System , Bioinformatics , DNA Classifying , Online Advertisement , Robotics , Robot Locomotion , Self Driving Car , User Behavior Analytics , Machine Perception …etc .

This single is not full enough to know about Machine Learning . This only covers the major and basic part of ML . Readers are advised to do some research more and practice on their own .