Traditional Programming VS Machine Learning Programming
Machine Learning has been introduced many years back but the kind of excitement it got from the last few years is unbelievable. Everyone is talking about Machine Learning but like with many things in the world it has been adopted with different meanings. Different people have different perceptions when it comes to the meaning of Machine Learning, Maybe all of them are correct in their own way.
But for a beginner it could be very confusing, this article is to ease that thought behind the basic definition of Machine Learning.
Before Understanding Machine Learning, let’s understand what is Traditional programming??
TRADITIONAL PROGRAMMING
Let’s say we have a function that takes 2 inputs (a and b) and returns the sum (a+b), so in traditional programming, we simply take input, put it into the function, and get the desired output.
This function could be any function, maybe function within function, maybe 100’s of function in one program but the basic block works this simple.
Machine Learning Programming
Now Let’s say we have input and also we have output, here the objective is to get the relationship between x and y.
let’s understand this first,
We have a function,
F(x)=y
here we have x, we have y and we need to approximate Function ‘F’.
Input X can also be called Features in Machine Learning World and Output Y can also be called Class Labels.
Let’s understand this by a Simple Example,
Imagine you got a job in Amazon as a Data Scientist and your manager asked you to build a Machine Learning Model which can predict whether a particular review given by an amazon customer is positive or negative.
Now before building the Machine Learning Model, let’s understand why it is important, Why reviews by Amazon customers important to Amazon?
Let’s say there is the product ‘ Blue Jeans’ on the amazon website and it has a total of 500 reviews, 400 people did not like the product, they gave bad reviews and only 100 people gave good reviews. Now the question is, “Will Amazon put this product on their first initial pages”?
The answer is “NO”, Amazon won’t put a product like this on their First initial pages. That is the reason why reviews from the Customers are very important for Amazon.
Coming back to Machine Learning Formulation…
DATASET EXAMPLE(WITHOUT LABELS)
Let’s say your manager gave you the data for the last 8 reviews given by the customers on a product “BLUE-JEANS”(Originally data is very very long than this, in real-world we might get 1 lakh reviews to come up with a model but for simplicity, I have taken only 8 reviews from 8 different people, some has given the positive review, some did not like the product.
Now a team in Amazon was assigned to label this data, whether it is positive or negative after reading the reviews.
DATASET EXAMPLE(WITHOUT LABELS)After manual labeling the reviews data will look something like this,
DATASET EXAMPLE(WITH LABELS)
DATASET EXAMPLE(WITH LABELS)Amazon gets thousands of reviews every second from all over the globe so if a Manual team will label every review, Amazon has to spend a lot of money and time to get these labels done.
SO DO MACHINE LEARNING HAS THE POTENTIAL TO LABEL THESE REVIEWS FOR US??
The answer is “YES”.
Now let’s connect the dots, we define machine learning programming as we have input and output both and we need to find the function F(X).
We need to find the relationship between Y and X.
Now we have X={Customer_id, Name, Product, Reviews} ( THIS IS OUR INPUT)
Now we have Y={Labels}
(THIS IS OUR OUTPUT)
We have input and output both and we need to find the Function F(X) which can give us the relationship between Y and X.
This means if tomorrow, Any random person reviews this jeans product then the Machine Learning System should be able to predict whether it’s a positive review or a negative review.
Thanks for reading…
