Machine Learning — Modeling the Future

In the above example, we saw how we learned from our experiences. There, we had 3 major elements — the data (here, how we fall down in each attempt), the model (here, the optimal model we created from our experience which we use in the next attempt) and mainly, the problem that we are trying to solve. In simple words, for every problem we get, we collect data from our experiences and find the optimal solution to it

Think that you are working on weather prediction. You are asked to predict tomorrows weather (if there will be a thunderstorm or not) and you are given weather information of the past 50 years (this is the data we use to train the machine. Let’s assume that it has hourly information about the rain, wind, and thunder for your region). Also, you have wind and rain sensors. Now, this is like a linear algebra problem on a large dataset, where X is the rain status, Y is the wind status and the unknown Z is the thunder status. Then, we train the machine using some machine learning algorithms (of which the core part is some mathematical function) to create a model (here, it may be a 3D graph) which we then use to find the current weather by giving the reading from the two sensors as input. Here, the machine acts like a guy who has experience in weather and knows to predict the weather based on the climatic changes. Woah! That’s the beauty of machine learning, making your own genie using mathematics, data and the ever-increasing computation power of machines.

So, why is it popular these days?

The idea of Machine Learning dates back to the 1900s, but it took more than half a century to come into reality. There are a lot of reasons we could point out. One major reason is the weak computational power of the processors that existed during that time. Also, there was no Internet as we have it right now nor any smartphones and this many users to collect the huge amount of data that is required for training. Apart from all these, there exists a huge investment required for the research purpose, which was then not available because of the lack of market for the products that could have been implemented using ML. These three limits the most important requirements for implementing machine learning — data, powerful machines, investment, and efficient algorithms.

In the modern world, ever since the coming of the Internet and smart devices, we have seen a huge advancement in the area of technology. This made smart technologies among the hottest market anyone could invest in. Due to the ever-growing demand, more and more investments came into the area of Machine Learning and Artificial Intelligence, which lead many companies and researchers to work in this area. Also, the huge set of open-source resources and the easy access to these resources via the Internet made it easy for young minds to try working in this domain. And the result? A whole new world of automated machines, that are smart enough to do several tasks that once only a man could do. We saw machines beating humans in playing alphaGo, helping us in analyzing the interstellar space to unfold mysteries, making self-driving cars, capturing beautiful images and many more that now has an influence on every human in one way or the other. Also, a whole lot of open-source Machine Learning libraries are available, which made it a whole lot easier for anyone with little knowledge to work in this domain. All these came together to make it the hottest domain of this decade.

What all could it do? or How could this technology help us?

Think of a situation where you want to predict how well your product will perform in the market. You have the data of all the previous products, their market performances and the parameters that affected its performance. Using this data, you could train a model to predict the current outcome based on the present parametric values and hence take strategies to improve its market performance. That’s not all. Read on…