Generalization to Specification (Machine Learning)

Even though Machine learning algorithms are helpful but getting to learn the Machine is very hectic process even with algorithms.

First we need to understand which problem we want to solve, and when it confirmed then as starting point we need to collect Data and the form in which the data should be.

Selecting General to specific or specific to general approach would be based on our problem statement.

For example: If we want our machine to recognize a Lion then we are talking here in generalize form because there are 100 of animals which have almost same elements like Lion so we must have to be specific here because even our machines are not good enough to learn the things in general/huge data form. Processor performance got down heavily during to initialize this data.

By applying “Specific” approach, first we write algorithm to collect specific Data and the form in which the Data should be. Then we will write algorithm to characterize that Data. When specific Data is collected then we can write algorithm to learn things for specific to general. For example in the case of “Lion” we have data that Lion has 4 legs, Body, A head and tail.

When we show our Machine “A Tiger” then it won’t consider the tiger as Lion but our algorithm will analyze elements of data from specified categories and then match those elements in Tiger. Obviously tiger have “checks on body” then algorithm again analyze data from that specific category which has “checks on Body” like Zebra, Tiger etc

Specific to general or general to specific algorithm sometime works as a “chain” and sometimes works in parallel.

Here could be more discussion on this topic but after this short discussion, we can say that “specific to general” algorithm is best for us because it has low error chances and more successful return rate.