Machine Learning: From a 5-year old’s perspective
Ever wondered how YouTube recommends your videos or how Facebook auto-tags your profile from a picture? Well, this is all the magic of Machine Learning. Today we are going to understand Machine Learning in layman terms. So what does the word “Machine Learning” means? The words ‘Machine Learning’ consists of two words “Machine” and “Learning”. The word ‘Machine’ stands for an apparatus that can perform any kind of task, like a computer, printer and so on. When we add the word ‘Learning’ after it, together it means how a machine learns. A machine can learn too like a human being and can be closely related to how a human learns.
So what is Machine Learning?
Machine learning (abbreviated as ML) is the ability to learn and improve from experience without being explicitly programmed. This can be explained with a very simple example.
For example, I take you to a pet shop which has 4 dogs and 4 cats and I teach you which of them is a dog and which of them is a cat. After this, I bring in a different breed of dog which was not present in those 8 animals. You can predict that it is a dog. So here I didn’t teach you that the other breed of the animal was a dog but you were still able to predict it as you were already taught previously. A machine works in a very similar manner. The 8 animals are equivalent to the ‘training data’ as it is used for training the machine. The other breed of dog is equivalent to the ‘testing data’ and is used for testing our machine.
What are the different ways a Machine can learn?
There are three types of Machine learning:
1. Supervised Machine Learning:
Supervised machine learning means when a machine learns in the presence of a trainer. The above was an example of supervised machine learning. In the above case, there might be times where you fail to correctly predict whether it is a dog or a cat. This leads to testing your accuracy about distinguishing between dogs and cats. Similarly, accuracy for the trained machine can be calculated too.
2. Unsupervised machine learning
Unlike supervised learning in unsupervised learning no teacher is present. It can be easily explained with the following example.
There are 4 dogs and 4 cats. Supposedly you have never seen a dog or a cat. So you can categorize them based on similarities, patterns, and differences. Thus we can categorize them in 2 groups: a group containing cats and the other containing dogs. Here you did not learn anything from before which means there was no training data.
3. Reinforcement learning
Reinforcement learning is about taking suitable actions to maximize reward in a particular situation. Let’s look at an example to understand it better.
Consider an example of a girl learning to cycle.
i) First, the girl will observe how we are cycling. We use two legs to paddle and try to keep balance. Now the girl will try to cycle.
ii) But soon she will realize that she has to balance the cycle to paddle properly. So now the girl tries to balance, failing most of the time but still determinant to keep the balance.
iii) Now the real challenge is paddling. There are so many things to keep in mind, like balancing the body weight, holding the handle firmly and look out for obstacles.
In the above example, the girl is an agent trying to manipulate the environment (cycling) and taking actions as she tries to go from one state to another. The girl gets a reward(say ice cream) for accomplishing part of the task(like paddling for a few seconds) and is being punished(not given ice cream) on not being able to cycle. This is the same way reinforcement learning is applied in machine learning.
What are some of the real-life applications of machine learning?
Facebook:
Facebook is the most widely used social media. Did you know that Facebook uses Machine Learning in almost every aspect? Whether your friend recommendations or auto-tagging of your friends in pictures Facebook uses Machine Learning.
Auto-Tagging of friends:
When a pic is uploaded on Facebook, a suggestion asking if you want to tag your friend in the pic appears. This is done by Facebook’s face detection and recognition algorithms which is based on the advanced deep learning neural network project Deepface.
Friend suggestions:
Machine learning is used by Facebook to suggest friends based on similar interests, area of living, etc.
Google:
Ever wondered how google assistant works? Well, it is based on machine learning and artificial intelligence. Some of the ways in which Google uses machine learning are as follows:
Google Music: It creates recommendations based on what kind of music you are listening to.
Ok, Google: This uses Speech recognition to interpret what you are asking it to do.
Google photos: Creates the categories of pictures you can search through. If we search ‘dogs’ it will show all images having dogs in it. It is a part of image recognition.
Suspicious activity detection from CCTVs:
Imagine a single person monitoring multiple video cameras. It will certainly be a difficult and boring job. So we train machines to do it. It captures video all the time and learns from it the normal activities of people like standing, running, etc. If any unusual activities occur like robbery, it warns the human attendants in real-time to prevent any mishap. And when such activities are reported and counted to be true, they help to improve the surveillance services.
Entertainment:
Do you know how Netflix recommends you movies? Let’s say users who watch movie A are likely to watch a movie B. This is perhaps the most well-known feature of Netflix. Netflix uses the watching history of other users with similar tastes to recommend what you may be most interested in watching next so that you stay engaged and continue your monthly subscription for more.
Healthcare:
Machine learning is used in hospitals to detect cancer. It helps doctors to make much more informed decisions by increasing the efficiency to recognize cancer and spot the cases where it is difficult for the doctors to identify.
Conclusion:
"….what we want is a machine that can learn from experience”
-Alan Turing
This can be regarded as one of the aims of machine learning. Most apps like Snapchat, Facebook, etc. that we use today are powered by ML. There are many fields where Machine Learning can still be applied. This will reduce human error and increase the efficiency of work. In the near future, machine learning will play a very important role in our day to day lives.
