Immanuel Manohar

A view of the person and profession


The increasing volumes of data currently being gathered as Big Data is to be analyzed using algorithms from signal processing, communication networks and statistics. The current focus of my research includes transportation of Big Data, efficient use of computing resources to solve data analysis tasks and finally development of models to study the data. This is achieved using a combination of techniques from the fields of communication, data engineering, statistics and machine learning. More details on my research statement can be found here.

Communication Networks:

My current research is in the field of communication networks. One branch of research is in developing a queuing model and the second branch is in combining computations with communications.

Queuing model:

Here, I'm trying to model a heterogenous network with high speed links using bulk transfer, bulk arrival queuing model with vacations.  I try to compare the performance modeling accuracy with existing queuing models. Also, try to answer the following questions:

1. Which nodes in the network are experiencing heavy loads?
2. If a new source has to join the network, How should the flows be redirected?
3. If there is a big data transfer operation hogging the bandwidth, what is the steady state loads in the nodes?
4. Is there an optimal routing and MAC strategy that could be developed using the same?

Computations and communications:

FoG nodes are of increasing interest in recent times. Here, we look into the distributed computation capability with communication restricted fog nodes. How can you distribute a data load?, how can you compensate for communication constraints? How can you combine communications with computations?



Other research interests are in image retrieval, matchine learning algorithms and methods of statistical analysis like PCA, ICA, and related factor analysis techniques.

The Communication Network Queue

Interesting Question: What is the best way to model a network? Which model is most useful? A Queue like the above or a flow based model?

Communication with Computation

Interesting Question: How to distribute computational load for huge volumes of data while considering communicaion load?

Factor Analysis

A bit controvorsial topic as you look for factors, you get factors which might be decieving. Here I don't talk about that but the different techniques for deriving statistically interesting factors that could be gleaned off huge time series data.

Image Retrieval

A simple and yet an elegant problem. Lots of people have looked into this and here's my perspective on what I have seen.

The Matrix QR Decomposition

While this is a cool decomposition, it's power and elegance is often overlooked in current research. People prefer Singular Value Decomposition (SVD) because of its simplicity, but how close does QR decomposition come to capture SVD details?

Big Data Transmission

I'm still learning this field, this is my research title for my PhD. I'll keep adding informaiton to this as and when I get to know more.