NATIONAL CONFERENCE ON CHALLENGES AND OPPORTUNITIES IN COMPUTER ENGINEERING – NCCOCE ‘17
17th and 18th FEBRUARY 2017

WORKSHOP ON DECISION MAKING UNDER UNCERTAINTY USING R PROGRAMMING LANGUAGE
(17th FEBRUARY 2017)

About the Workshop:
The future is ubiquitous with uncertainty. Yet the decisions are made in all aspects of life. Knowing and understanding the possible outcomes makes people smarter and their businesses competitive. For a long time, the ability to make decisions under uncertainty remained an art and a matter of wisdom. With the ability to collect and store colossal amounts of data, the intellect that is needed to make smart decisions is available. Simple tools are available to help learn from past experiences and make quantitative estimates of future outcomes. These tools are easily accessible to the executive of today’s businesses.

This workshop is an introduction to the underlying techniques in making smart decisions. These techniques are based on statistical techniques – specifically using Bayesian Model. Starting with demystifying concepts in statistics, various techniques with hands on tools that help in understanding how to evaluate future outcomes are carried out in the sessions. Participants will be exposed to a broad range of applications, including financial applications, medicine, court rooms, IT applications and SW management. The outcomes of this workshop form some of the basic underlying techniques of modern Machine Learning.

About the Speaker: Dr Satheesh S Kannegala, vice-Chairman of Computer Society of India Bangalore Chapter, and Chief Consultant – Technology Management, SecPod Technologies Sr.Technical Manager – HP, Sr. IT Specialist – IBM.

 

Dr. Sateesh S Kannegala is a Doctorate in Theoretical Physics, from University of Massachusetts, Amherst. He switched to the field of IT more than 20 years ago. He is currently a senior consultant providing data analytics services to clients that include Information Security, Air Pollution and Forecasting potential failures of vehicles. Dr. Sateesh S Kannegala has worked with Hewlett Packard and IBM in the past. He is on the advisory boards of many conferences. He is a member of the Board of Studies, Coimbatore Institute of Technology and has served as a member of thesis examination committee for M.Tech and Ph.D. students. Dr. Sateesh S Kannegala is currently serving as the elected Vice Chair of CSI Bangalore Chapter.

Session 1: 9:30 – 10:45 am:
The session began at 9:30 am.

Data Science: It is an interdisciplinary field about scientific methods, processes and systems to extract knowledge or insights from data in various forms, structured or unstructured, similar to Knowledge Discovery in Databases (KDD). It is new statistics, which employs techniques from Mathematics, Statistics, Information Science, Operations Research and Computer Science. Data Science Techniques are applicable to any domain. The major shift in techniques has come from increased computing capability and huge amounts of unstructured and structured data.

Examples: Voice recognition, Handwriting recognition (Image Processing), Fraud detection systems, Likelihood of Attack – Prediction, Analytics in Software Engineering, Fuel Economy Prediction, Health Care Analytics and Statistical Thinking in Legal Systems.

Machine Learning: It is a type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. Machine learning focuses on the development of computer programs that can change when exposed to new data. The process of machine learning is similar to that of data mining. Both systems search through data to look for patterns. However, instead of extracting data for human comprehension -- as is the case in data mining applications - machine learning uses that data to detect patterns in data and adjust program actions accordingly. Machine learning algorithms are often categorized as being supervised or unsupervised. Supervised algorithms can apply what has been learned in the past to new data. Unsupervised algorithms can draw inferences from datasets.

Session 2: 11:00 – 12:45 am:
The session began at 11:00 am.

Linear Regression: Linear regression is a statistical procedure for predicting the value of a dependent variable from an independent variable when the relationship between the variables can be described with a linear model. A linear regression equation can be written as Yp= mX + b, where Yp is the predicted value of the dependent variable, m is the slope of the regression line, and b is the Y-intercept of the regression line.
Cost Function: There could be error by using the defined hypothesis.

J (Theta) = (1 / 2 * m) Σ (Yi – h theta (Xi)) ^ 2

Gradient Descent: It is a first-order iterative optimization algorithm. To find a local minimum of a function using gradient descent, one takes steps proportional to the negative of the gradient (or of the approximate gradient) of the function at the current point.

Polynomial Regression: In statistics, polynomial regression is a form of linear regression in which the relationship between the independent variable x and the dependent variable y is modelled as an nth degree polynomial in x.

Session 3: 1:45 – 3:00 pm:
The session began at 1:45 pm.

Dr. Sateesh S Kannegala shifted his attention to the second class of problems: the classification type. Before delving into the actual fundamentals of the problem type, he gave us a multitude of examples to sufficiently define the problem type as well as to help understand its significance.
Further on, he took a specific classification problem, that of classifying a tumour as being benign or malignant. He showed us sample data sets of tumour sizes, which had to be classified into benign or malignant. He mathematically showed us how linear regression miserably failed to classify it and rather gave erroneous and potentially dangerous results.

It was at this juncture he introduced the concept of Logistic regression which used sigmoid function for regression calculation. He mathematically proved how logistic regression was far superior to linear regression in this application. He also went on to derive the cost function for logistic regression specific to this example and showed us how it can aptly generate a cost depending on the deviation of the result from actuality. This concluded the first half of session three.

In the second half of 3rd session, he started to talk about neural networks. He gave a very brief idea about what a neural network is and how it attempts to mimic the way neurons work in the human brain. He gave an insight into the gamut of its potential and told how it is used in deep net computations, autonomous vehicles and so on.

Session 4: 3:30 – 4:30 pm:
The session began at 3:30 pm

Bayesian Thinking: The Bayesian approach to learning is based on the subjective interpretation of probability. The value of the proportion p is unknown, and a person expresses his or her opinion about the uncertainty in the proportion by means of a probability distribution placed on a set of possible values of p.

•  Frequentists believe that probabilities are inherent objective property, independent of the person.
• All probabilities are subjective.

1.    Gained a broad fundamental understanding of the mathematical models and solution methods for decision making.
2.    An understanding of basics of Probability Calculus and their applications and common misconceptions.
3.    What Bayesian Statistics is and how it is different from traditional thinking and its applications.