Machine Learning - Basics

What is Machine Learning?

There are many perspectives on what machine learning is. But a well accepted definition is :

A Field of study that gives computers the ability to learn without being explicitly programmed - Arthur Samuel (1959)

A slightly complicated to understand but accurate one would be

A computer program is said to learn from experience E with respect to task T and some performance measure P, if its performance on T, as measured by P, improves with experience E - Tom Mitchell (1998)

Types of Machine Learning

There are most of the machine learning concepts some in one among any of the carders:

  • Supervised Learning -
  • Unsupervised Learning
  • Reinforcement Learning
  • Recommender Systems

Supervised Learning

Supervised learning is a learning technique used when we have a sample data with a given set of inputs and its corresponding ‘expected value’, we expect the computer to predict the ‘expected value’ when a new set of inputs is given to it.

Example:

We would take the most common example of Housing Price prediction. We would be given a list of ‘features’ of the house, area of the house and the cost of the house in that area, Our goal would be to predict the price of the house for an unseen house for which we know the height and width.

supervised

Now, we mark the area of the house (X-axis) along with its price (Y-axis) plotted in a graph as a (X) mark.

Now, what we do by the process called training, is to draw a function f(x) which maps x to y. In essence, y=f(x) is the function we will try to come up with so that for any new value of x, we can find the value of y by a process called predict. Here the list of features ‘x’ is called independent variance and the expected output ‘y’ is the dependent variable. Here the expected value ‘y’ is called independent variable.

If the independent variable is continuous valued, the type of supervised machine learning is called regression. The above one is a good example of regression based supervised machine learning.

The contradictory example is to predict the type of cancer, Malignant or Benign based on the size of the tumor. Here, the size of the tumor is the independent variable whereas, Cancer type is the dependent variable and can take one of the two values (Malignant or Benign) only. Hence this type of machine learning is called classification.

Classification

Unsupervised Learning

Unsupervised Learning is another category of Machine Learning where we would not be given the right set of answers. Instead, we give a set of data and ask the system, can you find a pattern in this data.

Example

One example of the unsupervised learning might be customer segmentation. We give give the system the customer segments and look for the system to segment out the customers without explicitly saying which segment any customer belongs to.

Unsupervised

Reinforcement Learning

Reinforcement Learning is a category of machine learning where the machine learns to perform a Task in an environment so as to maximize the reward over the long term.

A famous example might be the Alpha Go, developed by Google which bet the professional Go players.

Recommendation Systems

Recommendation systems are a class of machine learning which would predict the rating or preference of the user to a given item.

Example

Common examples are recommended products that are available in almost all online shopping sites such as Amazon, Flipkart, Ebay.

Recommendation

Disclaimer: Most part of the contents in this blog are from the Machine Learning course by Andrew Ng.