MACHINE LEARNING

 In our day to day life we're producing too much data. And hence we need to find a way to analyze, process and interpret this large amount of data. We need to find the way to operate on this large amount of data. A method which can make sense out of that data is known as MACHINE LEARNING. 

Machine Learning is most important because data is continuously increasing. Machine Learning improves the decision making, uncovers the hidden patterns and trends in data, solve complex problems and many more.

Arthur Samuel first define the term machine learning in 1959.

"A computer program said to learn from experience E with respect to some class of task T & performance measure P, if its performance at task in T as measured by P improves with experience E."

This definition is bit of tough to understand, lets make it understandable. 

Machine Learning is a science that allows computers to act or behave like humans. Computers improve their learning in the autonomous fashion over the time, by giving data and information in the form of observations and real-world interactions. 

There are so many definitions of machine learning. One if them is "Study of computer programs that leverage algorithms and statistical models to learn through inference and patterns without being explicitly programmed. "

Machine Learning includes some steps to operate:

1. Defining the problem that we are trying to solve.

2 .Data Collection.

3. Data preparing/ data processing.

4. Data exploration and analysis.

5. Building a machine learning model.

6. Evaluation of the model.

7. Prediction/Output.

In Machine Learning there are three major types:

1. SUPERVOSED MACHINE LEARNING

2. UNSUPERVISED MACHINE LEARNING 

3. REINFORCEMENT MACHINE LEARNING 

Lets understand one by one,

First one is Supervised machine learning.

In the technical words, supervised machine learning describes the problem that involves using the model to learn a mapping between a input examples and the target variables. 

As the named supervised, it works under supervision. Supervised machine learning works on labeled dataset. In this learning we train the machine using the well labeled data to find the correct data or correct answer. we can say that supervised learning is a task driven method or prediction method. The learning set is given and the objective is to form a description that can be used to predict unseen examples.

Using the set of variables, we generate a function that map input to desire output. The training process is continues until the model achieves the desire level of accuracy on the training data.

There are two main sorts of supervised learning problems. 

Classification:

Classification problem is when the output variable is a categorical, ex. "red", "blue", etc. Classification can defines to be group the output inside a class. If the algorithms tries to label input into two distinct classes then , this classification is called as binary classification. If the algorithm tries to label the input in more than two classes then that classification is called multiclass classification.

Regression:

Regression technique predicts a one output value using training data. It involves predicting a numeric label. Regression is used when the output variable is a real value, such as, "height", "weight", etc.

There are some algorithms which are regularly  used.

1.Linear Regression

2.Logistic Regression

3.Support Vector Machine

4.Navie-Bayes

5.K-nearest neighbor

6.Random Forest

Some applications of Supervised Machine Learning:

1. Bioinformatics

2. Quantitative Structure

3. Database Marketing

4. Handwriting Recognition

5. Pattern detection

6. Object Recognition in Computer Vision

7. Spam Detection

8. Optical Character Recognition 

Second type is Unsupervised machine learning

Unsupervised machine learning is the training of an algorithm using information that is not classified and not labeled and allowing the algorithm to act on that information without any supervision or guidance.

The main motive behind the unsupervised learning is to expose the machines to large number of varied data and allow it to learn and infer from the data.

Unsupervised learning problems are grouped into clustering and association.

Clustering :

A clustering problem is where you would like to discover the inherit groupings within the data, example grouping customers by purchasing behavior. 

In short, clustering means findings similarities from the data.

Association:

An association is where you want to discover rules that describes large portions of your data, such as people that buy x also trends to buy y.

Algorithms used in unsupervised learning :

1. K-means algorithms

2. Apriori algorithm

3. Expectation-maximization algorithm (EM)

4. Principle Component Algorithm (PCA)

Applications of Unsupervised Learning 

1. Humans behavior analysis

2. Social network analysis

3. Market segmentation of companies 

4. Organizing computer clusters based on similar event patterns and

    processes. 

Third one is Reinforcement Learning

Reinforcement Learning is allows machines to automatically determine the perfect behavior within a selected context, in order to maximize its performance.

Simple reward feedback is required for the agent to its behavior; which is called as reinforcement signal. 

Algorithms Used in reinforcement learning

1. Q-learning algorithm

2. State-action-reward-state-action algorithm (SARSA)

3. Deep Q network algorithm

4. Deep deterministic policy gradient algorithm (DDPG)

Applications of Reinforcement algorithm

1. Resource management in computer clusters

2. Traffic light control

3. Robotics

4. Web system configuration 

5. Personalized recommendations

6. Deep learning