Machine Learning Kya hai.

So in this blog, we will discuss what is Machine Learning. Is it really some disruptive technology? Why it has gained popularity over the past few years. This blog will help in getting familiar with it & basic terms used when talking about it. In the end, we will conclude what we have learned by answering these 3 questions.

So, what is Machine Learning?

Generic Tree to represent artificial intelligence and its branches.

As you might have guessed after seeing the above image Machine learning is a branch of Artificial Intelligence that needs a database. To get a better understanding, we need to know what is AI. My definition of AI is “with help of normal or complex algorithms to predict the result using machines”. To get a better understanding of artificial intelligence you can refer to my blog on the topic.

The definition of Machine Learning given by Wikipedia is “Machine learning (ML) is the study of computer algorithms that improve automatically through experience”. So, Machine Learning is just a study of a computer algorithm or more than that?

The answer is YES, it is much more than that. Let’s take an example to understand it better. Let’s assume you have a fever for the last 1–2 days and you want to know is it malaria, dengue, regular flu, or something more serious. So you decided to visit a doctor. The first thing they do is start asking you some general questions like:-

1. What did you eat in previous days?
2. From how many days you are having a fever?
3. Have you experienced any other symptoms like not getting hungry or having trouble in breathing.

After having above answers or more, they will recommend some tests to further narrowing down the cause. After getting the appropriate results, they will prescribe proper medications so you feel better.

Now think, what if those questions were asked by some chatbot. Doctors can directly look at the answers and tell you to get the relevant tests. It will save the doctor’s consultation time. Which will benefit more people on daily basis. The data are talking about in this scenario is the questions. Now you might have got some idea as to what is machine learning and the basic need for it to work.

Is it that disruptive?

And the answer here is also a YES. Here is a list of some of the numerous problems that machine learning has solved:-

1. Image Recognition:- Image recognition is the recognition of objects in images such as cancer in X-rays, or text in pictures. It is done on the basis of white and black pixels in black and white images and RGB pixels in colored images.
2. Speech Recognition:- This is another popular application of Machine Learning. It can translate speech into humans and machine-readable form. Some popular examples could be voice assistants like Alexa responding to your voice commands or voice google search.
3. Predictive analytics:- This is one of the underrated and most promising applications of Machine Learning. When there is random data, it classifies data into a bunch of groups. There can be rules set by an engineer on which these groups are made or machines it identifies themselves based on some characteristics. Real-world examples could be a prediction of a transaction whether it is fraudulent or legitimate.
4. Statistical arbitrage:- Based on Investopedia’s definition of arbitrage is “the purchase and sale of an asset in order to profit from a difference in the asset’s price between markets”. What it does is analysis of a market & large data sets of that market. After training on that data it helps us identify the real time arbitrage opportunities. Algorithmic Trading is a perfect example of this.

Despite being Machine learning in its early stages, there are many applications of it. So, now you might have figured out, why Machine learning is a disruptive technology. Now coming onto the next question.

Why has Machine learning gained popularity in the past couple of years?

There are multiple reasons behind it but the most important or could be called success pillars of Machine Learning. DATA, MODELS(specifically democratization of models) & DEVICES( efficient devices to compute data). So now we will learn a little more about these.

1. Data:- There is an abundance of data and in a different modality. What I mean from this statement is there is a huge amount of data and this data has multiple forms such as text, audio, video, structured, unstructured, and multilingual etc. There are platforms such as Pinterest(images), Facebook(images, text, videos, structured like name, age, interests, etc), Youtube(mostly videos), LinkedIn(text, structured data), Amazon(structured data, reviews), etc which generate terabytes sometimes petabytes of data daily. Previously when mobile phones and computers were not that common we were not generating the data at that big of speed. These data points are essential and most crucial in the success of Machine Learning.
2. Models:- The process of selection and training of the model is the job of a Machine learning engineer. Just remember this, Models are used to predict the outcome of Machine learning. And these models can range from simple to highly complex. And recently, some of these useful complex models were democratized by people. They also told us how to use them. This helped in boosting the applications of Machine learning in day to day activities. The more efficient your model is, the more accurate result you are going to get.
3. Computation:- To execute these highly complex algorithms on tons of data requires a high amount of computing power. Thanks to AWS, Microsoft Azure, Google Cloud Platform, etc cloud computing platforms, we can run our algorithms more efficiently by leveraging their computing power in a very cost-effective manner.

So these were the three success pillars of Machine Learning due to which it has gained a huge amount of popularity and success

Conclusion

Machine learning is a subset of Artificial Intelligence that needs data to predict the outcome. It groomed over the years and recently started showing it’s real potential. Many applications to it make a disruptive technology. There are many reasons behind the success of Machine Learning but the most important ones are Data, Models, and Devices. I hope you got the answers to all three questions in brief detail.

Thanks for reading..!!