Data Science, Machine Learning and Deep Learning. Are they all cut from the same cloth?

Machine Learning(ML) and Deep Learning(DL) comes under the broader umbrella of Artificial Intelligence(AI). The diagram shows that ML is a subset of AI and DL is a subset of ML, whereas Data Science is an intersection of Mathematics and AI along with all of its surpassing subsets. So, lets first give a look on what AI is?

Artificial Intelligence

AI is the capability of a machine to imitate intelligence or human behaviour. It would not be undue to call it one of the disruptive technologies. AI is an age old technology, but it gained popularity recently because today enormous amount of data is generated every single minute, and using the more advanced algorithms under AI, we can deal with that huge data size with an ease. AI basically is an umbrella of algorithms.Google’s Image Search Algorithm makes use of AI, as besides searching for the image, it also analyzes the information on it.

AI can be classified as follows:

• Narrow AI: Can only perform processing, e.g.-ALEXA.
• Artificial Generating Intelligence(AGI): Can perform multiple tasks as well as process them, e.g.-Robots.
• Super Intelligent AI: Outsmarted human Intelligence.

Tasks that constitutes AI are as follows:

• Problem Solving: Using efficient search algorithms as BFS, DFS, etc. No data or modelling is required.
• Knowledge Representation
• Reasoning
• Decision Making: Here’s where the intersection of AI with Data Science begins.
• Perception, Communication and Actuation: Perception using Computer Vision(CV) and speech Technology. Communication using Natural Language Processing. And Actuation involves Robotics and Reinforcement Learning(training an AI agent with data to take decisions)

MACHINE LEARNING

The most popular way of defining ML, goes like, it is the science of getting computers act without being explicitly programmed. Most powerful algorithms of AI, that are used to build intelligence, falls under ML.It makes use of implicit mathematics to understand complex patterns and also helps in predicting future.

Humans learn from experience, but Machines follow rules. ML makes computer learn from experiences. For computers or machines, experience is data. Some of the ML cloud Platforms are Microsoft Azure ML Studio, Google ML Platform, IBM Machine Learning, Amazon Machine Learning. ML comprises of 30% Hacking Skills, and the remaining 70% is the math behind it. ML can deal with large data size, even more than petabyte. We can say, “If AI is an inception, then on infusing algorithms to train it is ML.”

DEEP LEARNING

When there is a large amount of high-dimensional data available, then in order to learn very complex relationships between the input and output, a specific class of complex ML models and algorithms are used, collectively referred to ad Deep Learning(DL). It is inspired by the functionality of our brain cells, neurons, leading to the concept of Artificial Neural Networks. DL uses nested hierarchy of concepts at a large scale. Deep Learning algorithms are efficient enough to automatically identify the most important features for classification.

DATA SCIENCE

DATA SCIENCE is a multidisciplinary field, used for extracting information and knowledge from different forms of data available, and making use of different processes, algorithms, scientific strategies and systems in the process. It can be said that Data Science is a combination of Data Analysis and Machine Learning. It is the science of collecting, storing, processing, describing and modelling data. It comprises of Data, Domain Knowledge, Math and Stats as well as Hacking Skills.

Machine Learning(ML) vs Deep Learning(DL) vs Data Science

DL is a subset or higher evolution of ML. Feature Engineering is the technique to identify the right data from a given set of data. ML algorithms needs this feature engineering technique to identify the right or the required data from a given dataset. If wrong data is fed to the algorithm, then the performance is bad. But Feature Engineering becomes difficult for implications on images and texts, or any other forms of unstructured data, hence we use Deep Learning, as it can automatically identify the right data and can perform feature engineering on it. We can say that in this way DL algorithm learns how to learn. DL algorithms can be referred to as the black box algorithms, where the data goes in, decisions comes out, but the process in between input and output remains opaque.

ML is used in scenarios where we don’t want self-induced learning or interpretation, but only want to interpret the outcome for a certain set of input.

DL needs a large amount of data to perform well, but ML performs well on both large and small datasets.

ML usually breaks down a problem into subparts, solves them individually and then combine them to get the final results, unlike DL that solves the problem end to end.

DL algorithms takes a lot of time to train, whereas ML needs comparatively lesser time to train as compared to DL. This is because of the large number of parameters in DL.

On the other hand, Data Science deals with a series of steps including collecting, storing, processing, describing and modelling of data, encompassing the business, programming, statistics and communication domains.

Thus, it can be said that “Data Science, Machine Learning, and Deep Learning are all cut from the same cloth”.

Thanks for reading!