intelligence-544406_960_720

An Introduction to Machine learning and TensorFlow

All of us wish for an easy life where one doesn’t have to work hard but could lead a comfortable and luxurious life. But that seems like an impossible dream right? Alisha wished for something similar too, she was so annoyed with the amount of travelling she had to do every day and with the traffic that added to her trouble. She wished her car could drive on its own! Life could be so much easier that way. “Dream on!” one would say for that! But a person who knows about machine learning would know that it is as plausible to have a smart car in the near future as it is to have a smart phone in the present world. This is what Ben Stanley had to say about this in an interview, “Vehicles are intelligent today, they’re becoming a little bit more intuitive, but what we are seeing happening over the next 10 years is a notion that they’ll be able to do things by themselves.” Going by this we could say that by 2026 Alisha’s dream could come true!

Doctors can be replaced by software – 80% of them can. I’d much rather have a good machine learning system diagnose my disease than the median or average doctor.

But what is Machine learning and how does it make the impossible possible?

Machine learning isn’t a new concept, but it is gaining more prominence by the day. Machine learning is a process of analyzing data using algorithms to build a replica from the already existing data to make predictions. It is used to instruct the computer to accomplish things that are manually not possible to program in advance.

While smart cars could be the future accomplishment, what areas is it being used at present?

Data Mining: This is closely associated with machine learning and has various uses like revealing a fraud transaction of a credit card, understanding the customer purchasing practices and making predictions from it.

Text Analysis: This helps to segregate content such as mails, chats or any documents based on the text analysis. With the help of machine learning, computer perceives which mails to be classified as spam mails and diverts them to the spam folder.

Image Recognition: Basing on the previous tags made by the users, the algorithms recognize the specific face and makes suitable predictions. This is seen in popular sites like Facebook.

Robotic intelligence:  Scientists are using machine learning applications to make the robots smart. Unlike earlier where they could perform certain actions as programmed by the scientist, they would in future be able to learn and reason like the humans do. Smart cars being one of the examples.

Other areas: Machine learning is being implemented in various fields. For example, in sales, it gives live updates from the customers and vendors which help to avoid certain risks. In Marketing, it is used to give discounts or offers to its users based on their buying patterns. In Finance, it helps to make futuristic calculations about the returns that can be expected. And in Healthcare system, it is being used to prevent hospitalizations.

So, who’s providing a platform for Machine learning? Well, our very best friend, Google as usual comes to our aid with its second generation technology Tensorflow.

Before we delve into Tensorflow, there is a mention to be made of the branch of machine learning, which is deep learning. Some of the most successful deep learning methods involve artificial neural networks (ANN).  Incredible advancements in technology and abundance of data have made ANN mainstream.

Single-layer_feedforward_artificial_neural_network

Tensorflow is based on neural networks and was released by Google as an open source software library. It has already been tried and tested by Google themselves in few of their own products like Gmail to divert unwanted mails to the spam folder, in Google Photos which doesn’t just recognize the face but systematically sorts each element in the image to various categories even the text, in Google Search to predict the most searched for words and also in Google’s Speech recognition.

DisBelief was the first generation Google Brain’s machine learning system which was built in 2011 before TensorFlow and was used in Google search, Google voice search and Google Photos but since it had errors in neural networks, they wanted to make a much simpler and faster version which gave birth to TensorFlow.

Tensorflow is capable of running on various CPUs and GPUs. Be it Linux or Mac, or any mobile platform like the Android or Apple’s iOS. It instructs the neural networks to learn and reason like the humans by operating on multidimensional data arrays which are referred to as “tensors”.

A lot of companies have been using TensorFlow like EMC and have given quite satisfactory reviews about how efficient and less complicated it is compared to the other providers who offer Machine learning. TensorFlow could be the next BIG thing which is not only easy to use but can be ported between various hardwares too.

Murali Dodda is a Cloud Technology Specialist with over 15 years of experience. He graduated from the prestigious IIT Madras. Murali provides 'technology and business leadership' to startups and has overseen successful exits for several of them. He is currently leading a team of technologists at Bitmin, a hot new startup delivering cloud services. Murali uses his weekends to catch up on the latest developments in technology innovation, product development, and entrepreneurship domains. Being an enthusiastic blogger, he shares exciting developments & his experiences with designing & deploying cloud strategies through his blog. If you want an inside view of cloud deployment for real-world clients, follow this blog.

Leave a reply:

Your email address will not be published.

Site Footer