Let’s make Machines Learn!

“ Machine Learning for Noobies! ”

Many of you might have come across this hyped term ‘Machine Learning’ at some point of time ( at least that’s why you are here :P). We would be delving into some of the basic understanding of Machine Learning and what is it all about.

What is it?

Machine learning is a subset or a part of a much bigger concept known as Artificial Intelligence which specifically focuses on pattern recognition in various data available out there in the world.

More formally speaking —

Machine learning is the science of getting computers to act without being explicitly programmed.

Wait. What does that mean? Explain in noob’s term.

Andrew Ng solacing that you can study ML

So, Just imagine yourself (machine) being thrown into a vast desert (yes you have enough water to survive) and you need to escape out of here and reach habitable land (destination) using the most optimal path. But there exist millions of paths for escape as you can start in any direction but eventually (taking different times for different paths) you will end up escaping. So what just happened was, I told the machine (you) to reach some destination (escaping the desert) and the machine acted to fulfill the task. That’s Machine Learning! Woah you have rocket science in your hands now! Behold.

Searching for the right direction?

Now let’s consider some scenarios:

1. Before you start your journey, I give you some information about the different paths available (like which path would take a shorter time or maybe which one may lead to hell while another to the Cave of Wonder) and also the destination to reach. These pieces of information are known as ‘datasets’, which helps the machine to find patterns in the desert and find the optimal path to the destination. This is known as Supervised Learning.
2. But if you ended up in the middle of the desert without any such information or datasets then you would need to explore multiple paths in order to get out of the desert finding the most optimal path. This is Reinforcement Learning.
3. If the situation arises, that you are provided with prior information or dataset but the destination is not known then the machine is required to find patterns and connect the dots by itself in the given information and reach some destination. This is Unsupervised Learning.

Cave of Wonders

But Why?

Hey, but why do we need machine learning? Bruh, because it's everywhere! Right from where your reddened eyes checked the mobile early morning to the time you go to bed. You wake up checking your mails and find no spam mail lingering in the inbox. You see your emails being categorized into different labels automatically. You are sleeping in tents, enjoying holidays, and damn you forgot to inform your boss so you start writing a mail and see grammatical errors with suggestions popping up for spellchecks and basic responses. You have your breakfast and check social media for daily feed and find out that the suggested posts are similar to the ones you saw in the past few days. Being a millennial, it’s better to update your Instagram/Snapchat status about your studying and then see the filters being able to track your facial/eye movements. It’s lunchtime but with sheer suddenness, Zomato pops up a notification about the food which you beheld but didn’t order about a week ago. Damn! You scream, “Hey Google, open Zomato!”.

What is Deep Learning?

Well, here comes the more complex part. ‘Deep Learning’ is many times used interchangeably with ML but it is a further subset of ML, used to solve more complex problems.

Deep learning uses neural networks, which helps in a better understanding of the available data. It has various algorithms and different neural network architectures available.

This is how messy neural networks can look —

Deep Neural Network

But, it’s better to start off here —

Neural Network for Babies

Why Deep Learning?

Andrew Ng says -

I think AI is akin to building a rocket ship. You need a huge engine and a lot of fuel. If you have a large engine and a tiny amount of fuel, you won’t make it to orbit. If you have a tiny engine and a ton of fuel, you can’t even lift off. To build a rocket you need a huge engine and a lot of fuel. The analogy to deep learning is that the rocket engine is the deep learning models and the fuel is the huge amounts of data we can feed to these algorithms.

As we advance more towards development, the amount of data available with us increases exponentially, so it starts showing vulnerable results with old algorithms but at the same time neural network does a great job in handling such an enormous amount of data.

Peace ✌

Reference

What are the types of machine learning?