With data growing at an enormous speed, the techniques required to gain fruitful information from it have also grown substantially. Machine learning, is one of the techniques which provides with various methods which could be used to get insights about the data. Linear regression, Linear Classification, Neural Networks are some the types of machine learning techniques.

Starting with Machine Learning,the primary step is that the data should be cleaned so as to remove any kind of disturbances. Sometimes there are null values in the data, so to fill in, various techniques could be employed such as taking the average of row or column or performing linear regression on column. After this, data has to be divided into 3 sections: one for learning , another for validation and finally for testing. The data should be then normalised so as to remove any kind of deviation. After these steps are done, the data is ready for use.

If you don’t have the knowledge of the techniques which could be used on the data, then source which could be of help to you is the “MACHINE LEARNING” course by Andrew Ng from Stanford University. He has his video lectures on Coursera which give ample detail about the implementation of the same. So depending upon the type of output which is desired appropriate technique should be chosen.

1) In Linear Regression, Gradient Descent algorithm could be used to get the value of parameter Theta accurately, which will further lead minimum deviation from the actual data. To calculate the difference between the calculated and observed value, one must use Squared Mean Difference. Also one should see to it that they don’t over fit the linear line which is generated.

2) While using Neural networks, the concept of neurons is highlighted. Neurons in brain send and receive signals which help in information transmission. Back propagation is the algorithm which is of prime importance while making a Neural network. It is computationally demanding, that means it requires a lot of time to process and also a good processor depending on the number of hidden input layers present in it.

With the implementation of these steps you are ready to make your machine learn!!

Here’s the link to MACHINE LEARNING course:

https://www.coursera.org/course/ml

Very nicely written. Keep it up.

What I conclude from your article is that if hardware is not a problem then Artifical Neural network becomes the indisputable choice for the machine learning implementation. isnt’ it ?

What all ANN implementations available for ML. What are the benefits of one implementation over another and what factors will derive to make a choice among them.