Data Science vs Machine Learning vs Deep Learning : A Case Study Approach
CASE STUDY
Mr DS worked for 20 years in a Tour and Travel agency and made a lot of money. He decided to start his own Tour agency with one of the core product based on image recognition. This product would automatically detect monuments in a picture uploaded by the user and tag the image with that monument. He hired two of his friends as employees, Mr ML and Mr DL for building the product. Using the experience from his past job, Mr DS collected several images of monuments from different parts of the country and stored them efficiently in a database.
He then asked his 2 employees to build a model which would automatically identify the monument. Mr ML, a very hard working person, took the images and tried several ways to extract features, like edge detection, image sharpening and other image processing techniques. He then passed these extracted features through a model, which would then identify the monument, as shown below
Mr DL on the other hand was quite a lazy person and decided that he would let the model learn the features instead of manually working on it. He directly passed the image through the model and trained it to identify the monument.
Unfortunately, within 6 months the startup closed for lack of future ideas, inspite of the good work of all 3 of them.
STORY EXPLAINED
Mr DS stands for Data science, incase you did not figure it out. As in the story, data science is an umbrella of several tasks, which include: translating a business problem to a data science one, collecting data, storing it, processing it, describe the data and finally building a model to understand the underlying relations and even make predictions on new data.
Machine learning, represented by Mr ML, deals mainly with the modeling aspect of data science. In ML, we try to estimate the relationship between the input variables and the output using a large family of complex functions. These functions(also called model), for a new input point will predict an output. Thus it intersects with Data science. At the same time, we cannot call machine learning a complete subset of data science, as it includes topics like algorithmic efficiency, which concern more with computer science than with data science.
Deep Learning, represented by Mr DL, does the same thing as ML and hence is a subset of ML. The major difference between ML and DL, as also shown in the story, is that in ML we create the features and then pass it to the model, whereas in DL, we allow the model to create the features by itself. For example in the image example shown above, operations like image sharpening and edge detection are done by certain filters(matrix), as shown below.
In DL however, instead of using these specific filters, we allow the model to determine the filter coefficient.
I hope I was able to explain the story and the difference between DS, ML and AI. Cheers.
