ISME

Explore - Experience - Excel

Machine Learning and Neural Network: An Introduction to Beginners

Machine Learning and Neural Network: An Introduction to Beginners
Prof. (Dr.) S. Shyam Prasad
Introduction
In the present era of big data and hypercompetition a plethora of tools is being used to predict the future as accurately as possible to gain competitive advantage. One often comes across subjects such as Machine Learning and Neural Network amongst many other. The intention of this small write up is to introduce the reader to these terms and particularly the students to make them aware so that they may prepare themselves for deeper study or a career in them if they find it interesting. The institutes can also think of introducing short-term courses in these subjects.
Machine Learning
We have been using machine learning many times a day without realizing it. A simple example would be Google search. Every time we do a Google search, the search engine works so well because Google’s machine learning software guesses our search intention and accordingly ranks the pages. Even the email spam filter that separates out the chaff and saves us the time and effort in handling emails is an example of machine learning. At a higher level, getting robots to drive a car or tidy up the house are also examples of machine learning. Scientists are hopeful that progress in this direction can be made through learning algorithms called neural networks. Neural networks imitate our brain and resemble its working. Discussion on neural networks is done later in this write up.
On going through the literature of machine learning, one comes across two definitions of machine learning. An earlier definition by Arthur Samuel described machine learning as: “the field of study that gives computers the ability to learn without being explicitly programmed.”
A formal and a modern definition is given by Tom Mitchell and the definition is:  “A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.”
Example: playing chess.
E = the experience of playing many games of chess
T = the task of playing chess.
P = the probability that the program will win the next game.
Broadly, machine learning problems can be classified as: Supervised learning and Unsupervised learning.
Supervised Learning
In supervised learning, we know both the input variable say ‘x’ and the correct output say ‘y’ and we also have the knowledge that there is a relationship between the input and the output. These problems can be of two types: regression and classification problems. In a regression problem, we try to predict results within a continuous output. In other words, we try to map input variables to some continuous function. Instead, in a classification problem, we try to predict results in a discrete output. In other words, we try to map input variables into discrete categories.
Unsupervised Learning
In unsupervised learning, we have the input variable say ‘x’ but have little or no idea about correct output. All that we do is derive a structure from that data where we don’t necessarily know the effect of the variables.  This is done by clustering the data based on relationships among the variables in the data.
Further explanation or discussion would require mathematics and so we dwell no further in this direction. Readers interested in the topic can choose a simple course from MOOC for further reading. However, before moving, let us understand the reasons for machine learning being so prevalent today. In fact, mac
hine learning is a field that has grown out of the field of artificial intelligence, commonly known as AI. In the field of AI, the scientist while attempting to build intelligent machines found that they can programme a machine to do some basic things such as how to find the shortest path from A to B but did not know how to write AI programmes to do more interesting and complex things. Scientists realized that the only way to do these things was to have the machine learn to do it by itself. So, machine learning was developed as a new capability for computers and today it touches many segments of industry and basic science. Range of problems that machine learning touches is far and wide such as robotics, computational biology and tons of things in Silicon Valley. For example, when every time one goes to Amazon or Netflix or iTunes Genius, it recommends the movies or products and music to buy and that’s machine learning algorithm at work. However, there are about million users; one cannot write million different programmes for million users. The only solution that one can think of is that the software has to learn by itself to customize the preferences.
In a slightly dated article accessed over internet that listed top IT skills, at the top of this list of the twelve most desirable IT skills was machine learning. At least in US, a number of recruiters contact universities and enquire about graduating machine learning students.  Apparently, there seems to exist a vast, unfulfilled demand for this skill set, and this is a great time to learn about machine learning.
Neural Networks
Neural Network is also a machine learning algorithm. Neutral network was actually an old idea, but had fallen out of favor for a while. But today, it is the state of the art technique for many different machine learning problems. Neural Networks are actually very effective state of the art technique for modern day machine learning applications and get them to work well on problems.
Conclusion
Machine learning is a method of data analysis.  The specialty of machine learning algorithm is that on being used repeatedly it learns from data.  It allows computers to find hidden insights without being explicitly programmed as where to look. As mentioned earlier, growing volumes and varieties of available data, increasing power of computation and reducing cost of computation and more and cheaper data storage facilities have all made machine learning very popular.
Machine learning is now used in market segmentation, predicting consumer behaviour, employees’ growth and career more accurately, segregating more potential patients for certain diseases and many more. It emerges from above that the scope of machine learning seems to be unlimited.

Armed with above knowledge, one is in a position to understand the terms machine learning and Neural Network. In general, there is wide scope for people with good knowledge of mathematics and particularly IT and electrical engineers to cash in on this opportunity. Management knowledge would be an added advantage. 
X