Yantra Budhi -- Basic Introduction of Machine Learning

Machine Learning basically has two definitions. They are:

1. Arthur Samuel -: "The field of study that gives computers the ability to learn without being explicitly programmed." This can be considered as an older, informal definition.
2. Tom Mitchell explains how the computer program became a Machine learning one here -: "A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E."

The second definition is considered in the modern world as it is having more clarity with the parameters which is more quantitative.


Now basically the question falls that how the machine is learning the things!


In an overview, it is being again subdivided as

1. Supervised Learning
2. Unsupervised Learning

From the name itself, some things are clear that one is by someone supervising the phenomenon of learning and the other there is no one to supervise the learning.

1. Supervised Learning:

Now, normally one will think who is supervising. The machine itself is supervising but the data supervises machine to calculate and find the result. We have concrete data which by some equations or classification parameters we can reach to the result.

Again this is being subdivided as:

• Regression :
• Predicts a continuous-valued output. Many parameters will be there to find out a result.
• Eg:- Given a picture find age.
• Classification :
• Separate the given data to discrete value of output.
• Eg:- Given many pictures classify them to different age categories. 
• 
  

2. Unsupervised Learning:

Now, what is unsupervised. Complete mystery of finding the result. Yes, it is not having any feedback to correct means no input data early available to supervise. The data which are available to work on as well as the scenario to separate and find a result from the data is new in this case. We can conclude Unsupervised Learning as if we have some data and we can separate it into meaningful information without any external input.

• Clustering:
• Given data and separate finding factors into clusters.
• Eg:-Astronomical data to find similarities and cluster it.
• Non-Clustering:
• Given data instant of seperate time period or dimension and classify within from the data to find the difference.
• Eg:- from two videos of same two person in an area at different angle directed with seperate camera and from this data distinguish the person and label in the video without any previous learning data.