2006  2007
Stochastic Processes and Applications
August 2006  July 2007 

"Stochastic" is a Greek word which means "random" or "chance". Stochastic Analysis deals with models which involve uncertainties or randomness. Uncertainty, complexity and dynamism have been continuing challenges to our understanding and control of our physical environment. Everyday we encounter signals which cannot be modeled exactly by an analytic expression or in a deterministic way. Examples of such signals are ordinary speech waveforms, seismological signals, biological signals, temperature histories, communication signals etc. In manufacturing domain no machine is totally reliable. Every machine fails at some random time. Thus in a typical manufacturing system which involves a large number of machines, the total number of machines at any time cannot be determined in a deterministic way. In a market driven economy, the stock market is volatile, the interest rates fluctuate in a random fashion. One can give any number of examples from our daily life events where uncertainty prevails in an essential way. This gives us the realization that many real life phenomena require the analysis of a system in a probabilistic setting rather than in a deterministic setting. Thus stochastic models are becoming increasingly important for understanding or making performance evaluation of complex systems in a broad spectrum of fields.
Under the current thematic programme we first plan to carry out training and research in mathematical finance. Mathematical finance is an interdisciplinary area. It lies at the crossroad of probability theory, partial differential equations, numerical methods, statistical analysis and economics. We plan to organize a series of lectures on Mathematical Finance throughout the year. We also plan to organize a workshop for researchers from the academic institutions and practitioners from the finance industries followed by a conference. In this manner, we wish to create a big group in this area across the country which would grow to meet the challenges thrust on us by the changing socioeconomicpolitical scenario in our country.
Stochastic processes have played a significant role in various engineering disciplines like power systems, robotics, automotive technology, signal processing, manufacturing systems, semiconductor manufacturing, communication networks, wireless networks etc. Among the above engineering applications of stochastic processes, we plan to concentrate on communication networks where there is lot of current interest. This will form the second part of the thematic programme.
In communication networks, unpredictability and randomness arise for several reasons. For one, connections come in and leave randomly. At the time of connection establishment, circuit switched networks typically need to find sufficient resources for the connection in the absence of which, a connection is refused. For packet switched networks such as the internet, even though the above constraint is not there, however, each packet is routed individually by the switches based on current information available and can take a different path. Also, packets can be dropped if sufficient resources are not available. In most networks (in particular, packet switched), the packets that individual connections pump into the network are of varying sizes (one packet may have a different size than the other in each connection) and thus each packet holds the network for a varying (random) amount of time. Moreover, the links can fail randomly and so reliability of the medium is also an issue. From the perspective of the user, the total time needed to transmit, say a file, should be the least possible. Whereas processing, transmission and propagation delays of packets are not significant in general, queuing delays are. The latter depend on the size of packets and also the number of connections operational at a given time. Stochastic processes are thus crucially used in the design, analysis and control of networks. Control in networks can be broadly classified under four heads  admission control, routing, flow and congestion control, and resource allocation. Designing good control strategies requires a good knowledge of (the above mentioned) topics such as stochastic control (in particular under partial information), parameter estimation, simulation based optimization, queuing theory, learning theory etc., for all of which stochastic processes form the key ingredient.
Another area where stochastic processes have important applications is in the area of neuroscience. This will comprise the third part of the thematic programme. The firing of neurons can be modeled as a first passage time problem from stochastic processes. The output spike train is a point stochastic process. Analysis of EEG signals from the brain makes intensive use of stochastic processes, in particular, vector autoregressive processes. Among the awesome repertoire of tasks that the human brain can accomplish, one of the more fascinating ones is how the electrical activity of millions of brain cells (neurons) is translated into precise sequences of movements. One of the greatest challenges in applied neuroscience is to build prosthetic limbs controlled by neural signals from the brain. The ultimate goal is to provide paralytic patients and amputees with the means to move and communicate by controlling the prosthetic device using brain activity. Scientists and engineers are slowly getting closer to building such devices thanks to studies revealing a strong connection between the activity of neurons in the brain's cerebral cortex and the movements of limbs. To realize the above goal of building prosthetic limbs, one tool which plays a critical role is the theory of stochastic processes.
As part of the thematic programme, we plan to organize seminars, compact courses, workshops and conferences, etc.
