MACHINE LEARNING AN INTRODUCTION AND ITS APPLICATION 

• Machine Learning an introduction
• Definition of Machine learning
• Why do we use Machine Learning?
• Summary of Machine Learning
• Examples of Applications
• Types of Machine Learning 

Machine Learning an Introduction

When we hear the word Machine Learning, we imagine a picture of the robot. Machine Learning is already with us and used as a futuristic fantasy. It has been with us for decades in some specialized applications.

The spam filter is technically qualified as Machine Learning. You have to flag an email as spam anymore. 

Where does Machine Learning start and end? What does a machine do to learn? If I download a copy of Google, has my computer really learned something and suddenly smarter?

We are very friendly with emails, and understand machine learning using the example of Spam email. The spam filter is a Machine Learning program. The example of spam emails is flagged by a user and an example of regular email is non-spam or "ham".  

Definition of Machine learning

“Machine Learning is the study of programming computers that can learn from data.”

Machine learning is a part of Data Science and paying an important role to be a Data Scientist.

If you download a copy of Google and your computer is not better at any task with a lot of more data suddenly, downloading a copy is not Machine Learning.

Why do we use Machine Learning?

Using traditional programming techniques how you can write a spam filter.

You might notice some words like Free, amazing, Credit card, 4U come up with a lot in the text line. You would write an algorithm to detect flag emails as spasm for each of the patterns.

In the case of machine learning techniques that filter spam words and phrases learning automatically detects the pattern of spasm words. This program is shorter, easier, and accurate than the traditional program.

If spammers know that the '4U' word is blocked, they might start writing 'For U’ instead. The traditional technique cannot find out the spasm in this rule but machine learning automatically detects it.

In another area, Machine Learning is not used for too complex traditional approaches or unknown algorithms. For example, speech recognition. If you want to write a program with distinguishing the words “one” and “two” you notice the word 'two' starts with a high-pitch sound ('T').  So you have to hardcore an algorithm that measures high-pitch sound intensity. But this technique cannot be scale thousand of words spoken by millions of people in a noisy environment with different languages. The best solution is Machine learning that learns algorithms by itself from given many recordings of each word.  

For example, once a spam filter is trained, it can be easily inspected the list of words and are the best predictors of spam. This will also express new trends and lead to a good understanding of a problem. When you dig a large amount of data by applying Machine learning can discover a pattern is called Data mining.

Summary of Machine Learning

We require a lot of long list rules to solve a problem. But a Machine learning algorithm can simplify the code and perform better than the traditional approach.

The traditional approach cannot yield a good solution for complex problems. So Machine learning techniques can be a good solution. A machine learning system can adjust to new data. The machine learning system is getting information to solve complex problems and large amounts of data.

Examples of Applications

Some examples of Machine learning are given below. We can automatically classify them after analyzing images of the product.

1. The image classification performs using convolution neural networks. This method helps to detect tumors in brain scans. Here each pixel of the image is classified to determine the exact location and shape of tumors. This is called semantic segmentation.
2. The news articles classify automatically using natural language processing (NLP) and text classification using recurrent neural networks (RNNs), CNNs, or Transformers. NLP tools also flag offensive comments on discussing forums automatically. NLP summarizes long text documents automatically.
3. Creating a chatbot or a personal assistant - Chatbot and personal assistant involve many NLP components, and including natural language understanding (NLU) and question-answering module.
4. We can forecast the revenue of any organization for next year, based on many performance metrics using any regression model. This is called a regression task such as a Linear Regression or Polynomial Regression model, a regression SVM, a regression Random Forest, or an artificial neural network.  If you take into account sequences of past performance, you may use RNNs, CNNs, or Transformers.
5. The application of voice commands is called speech recognition. The processing of audio samples is long and complex sequences that using RNNs, CNNs, or Transformers.
6. Detecting credit card fraud is anomaly detection - You can design a different marketing strategy based on the client's purchase for each segment. Clustering is the representation of a complex, high-dimensional dataset in a clear and insightful diagram.  Data Visualization is often involving dimensionality reduction techniques.
7. Recommending a product is based on the interest of a client's past purchases. This is called a recommender system. 
8. Reinforcement Learning helps to build an intelligent bot for games which is a branch of machine learning. A bot may get a reward whenever the player loses life within a given environment.

This task gives the feeling of incredible complexity that machine learning can tackle.

Types of Machine Learning 

The different types of Machine Learning systems are useful to classify in broad categories. These are as follows:

• supervision supervised 
• unsupervised 
• semisupervised
• Reinforcement Learning
• Online versus batch learning
• Instance-based versus model-based learning