I consider myself strong in algorithms, data structures and programming. From last few years, I was interested to be an expert in Deep Learning Technologies.

My initial understanding was that to be a good consultant in Deep Learning, I need to learn too many things to make the application work, write monstrous code using a lot of APIs, understand lot of mathematics (calculus, algebra and probability) to grasp the concepts. But my passion to master this new technology made me jump into it a year back. Then there was no looking back.

I have just ended a semester teaching masters and Ph.D. students a course on “Deep Learning”.  Last week I was able to successfully lead an International workshop on “Deep Learning” and able to test that how quickly people can kickstart their journey of Deep Learning Technologies. Also, I was also able to consult few startups on Deep Learning.

I was wrong in my understanding of the complexity of Deep Learning Technologies, because it turned out that it is very easy to learn and master these technologies. Very little math is required. Wonderful courses, Tutorials and help is available online. Deep learning platforms hide the complexity from the programmer, that results in a simpler code. You will be surprised to see the size of the code of various Deep Learning projects, if you compare them with classical applications and projects of computer Science. Soon, we will have tools where users will be able to play with various hyperparameters of any application they want to develop; without worrying about the inner details of the black box. So, I will appeal to our young generation in the colleges and universities to grasp this opportunity for their own benefit and start learning Machine Learning Technologies.

What I learnt from my learning journey is the immense power of Deep Learning Technologies. It has the hidden and unexplored potential in many fields of life.  From the technical point of view, it only looks like a function approximator. But, people are awestruck when they found that it can do real time language translation and do real time driverless vehicle maneuvering with simple weight optimization using gradient descent methods. Now, People are exploiting it for all possible applications in every walk of life. For me the real use lies in Health Care. If the developer and researcher community along with the industry can spent more time in perfecting the Deep Learning techniques for detection, analysis and diagnosis of diseases, then it will be a game changer. It can be another revolution which can further increase the life expectancy on earth. Early results have already started coming in various Research journals and articles. To take these innovations to the level so that common man has access to it, will take time and efforts.

It is natural that companies are spending more resources in the applications, which have commercial benefits to them in the short and long term. For example: targeted advertising, product recommendations, improving the customer experience at point of sale websites etc.  Now, the trend has started taking a new turn and many start-ups are using these technologies for online education, precision agriculture, aids for specially-abled people, policy making of states and nations, finding patterns in financial transactions, security and maintenance logs of automation systems etc.

As the more advanced areas of Autoencoders and Deep Generative Models are taking shape, it will be exciting to be a partner in this journey of Deep learning.

Already the number of datasets that have been made available as open-source are increasing. But, the expectations of developers are lot more from institutions, Govts and Industry. Early these organizations shed their resistance to make the data available, better it will be for the developers and researchers to come out with tangible solutions to many of the world’s problems. It is in the interest of everybody that we start sharing the data. Nobody should worry about the privacy part as there are techniques and methods to anonymize the data and that takes care of the privacy part without any trade-off with its utility.

This article first appeared in the TOI