Machine Learning: Fundamental Concepts

Machine Learning is a subject which deals with an intelligent  system   which can prove it can learn by providing expected evidence when it is presented to a testing purpose .

What is learning ?

Learning is a process which grows a special knowledge about a thing and a general idea about its behavior like a natural choice .

 Machine Learning?

the goal of machine learning is to build a system or algorithm   which can automatically works like human brain with an expected amount of accuracy  that proves its really  a interesting  matter.

The objective of the current trends in machine learning   is to utilize the proper mechanism of human brain perception thinking and deep architecture  and  provide a unlimited learning capability with a proper structural way  that reduces the working complexity and predictability  with a relative very nano time.

Some Practical examples are discussed there :

House Keeper Assistant

    Suppose the main door of the house is automatically closed without any help of human . To protected security it valid for  only house members. So how the door will opened or closed . So, There need a two-eyed person or one or more artificial eye which can recognizes  right persons . If there is way which can capture the picture of the man who is willing to enter the house and send the most important person who can identify the people or not . That house person then may send a message which atomically can to the persons and the machine . Then the machine matches the secret  message to the message that also come to the machine , if matches then the door is automatically often and the person enter the room. Here  the accuracy  of the system is 100% without considering  the security issue.

But the camera embedded machine can not identify the man the next without the new message or the captured image. Even if there is some special password for the family members only , that may be interesting approach but if the door just atomically on or off and a a huge entrance  happens occasionally  or everyday  , then the family members get boring someday .

So,   if there is some facility to the system that can detect people who is known or unknown according to the previous history , that really cools ,which reduces the notification to the house peoples. How can we build such a system? What are the requirements?



The requirements are

1. Face Detection

2. Eye Detection

3. Total Body Detection .

Here we think , the people of the house wants to setup a  Face Detection Machine in front of the door .

They think  , how it possible

1. It is quick impractical to exact machine of face to face , i,e, the pixel representation to pixel representation as there is a minimum amount of deviations occurs on every looking, emotional effects or something. that results  the system try to one-to one corresponds which gives very low accuracy.

there is no way?

if there is many pictures of a man and we try to match one to one pictures and checks who is the person . its really a bad way as  the internal architecture have a great similarity .

So if we avoid the exact data that represented  the original face and take the features like eye colors, radius of eye, skin color, the width of the fore head, the mouth pattern etc. And try to learn a one concluding way that maps the variations of this particular feature value obtained from the similar image of a single persons , and obtain a quick decision about the people which with an smart decision that  the person is record or known , that  towards a better solution , like human can do.

So, why need multiple facial image of a single person if we expects better accuracy ,

Human can image some variations  even there is a single image of a man with different aspects  according to the experience or inherent capability . But machine can’t do that .

And We can’t believe  machine as like we believe a human being . SO our believes is to make a perfect decision which come through a probabilistic assumption or   exact  matching , needs to give a strong reference to machine . That is not enough that accuracy always depends how much inputs we give. Its not true most of the case ,

In this context , if we give some facial image of man’s childhood ,some with sun-glass  and some with the current attitude . That ensure that some interesting this that may know the system but the purpose of detecting himself is totally depended on the current view of the person. So, relation between the data is an important things.










Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s