M.Tech. Projects

[Project Topics from Jan, 2016][Project Topics from Jan, 2015] | [Project Topics from Jan, 2014]

Proposed M.Tech. Computational Science Project Proposals

Jan 2017


1. M.Tech projects in Computational & Statistical Physics Lab (Murugesan Venkatapathi)

CSPL is open to one or two M.Tech students joining the group. The potential areas/topics for final projects are:

      • Numerical Linear Algebra: A rigorous theoretical and numerical analysis of preconditioners used for linear solvers
      • Partial Differential Equations: Mapping of separable solutions of a hyperbolic PDE to coupled ordinary differential equations of some simple dynamical systems, for effective physical interpretation of phenomena.
      • Computational Electromagnetics: Formulation and Implementation of a multiscale Discrete Dipole Approximation (DDA)
      • Any other ‘specific and suitable’ topic the student proposes which involves ‘Matrix Algebra/Analysis’ or application of ‘Statistical models’

Students with a background preparation in corresponding areas and a srong motivation to continue research beyond M.Tech will be preferred.


2. Medical Imaging Group (Phaneendra Yalavarthy)

Only those students who have strong inclination towards research should pursue these projects. The expected outcome of the M.Tech. project will be a journal paper.

  • Novel Computational Methods for Quantitative Susceptibility Mapping (QSM): QSM has become a standard Magnetic Resonance (MR) Imaging protocol for assessing the iron content and changes in venous oxygen in Brain. Reconstruction of magnetic susceptibility distribution requires solving an inverse problem and typically involves advanced computational methods. The project aims to look at the computational aspects of the inverse problem and propose a novel image reconstruction method for QSM.
  • Acoustic Wave Propagation Modeling for Photoacoustic Tomography (PAT): The photoacoustic imaging combines the optical and ultrasound imaging modalities to provide better structural and physiological details about the tissue under investigation. The PAT reconstruction problem is an initial value problem with limited data, which uses the acoustic wave propagation as a forward model. The aim of this project will be to look at available acoustic wave propagation models and assess their suitability for various tissue sizes/types to bring out the advantage and limitation of each of them.

3. DREAM:Lab (Yogesh Simmhan)

The following project topics are open to students strongly motivated towards research outcomes.  Students should be willing to read research papers, understand new topics, develop research prototypes, and perform experiments. Taking the DS265 Scalable Systems for Data Science course in Jan 2017 is a must.

  • Analytics over Dynamic Graphs: This project will examine algorithms and analytics that are suitable for large graphs that have a dynamic or time-varying nature. Such graphs include social networks, traffic and IoT networks, knowledge graphs, etc. where both the structure of the graph as well as its properties can vary. Such analytics may need to operate on the changing graph in real-time, or over offline data that has been collected. Analytics can include community detection and evolution, time-varying shortest paths, but many more remain to be explored over emerging datasets. When designing novel dynamic graph analytics, distributed and Big Data programming models will need to be leveraged to implement them in a scalable manner. Students interested in this topic will require strong skills in algorithms, and be willing to learn distributed and graph algorithms, and graph theory.
  • Big Data Platforms for Dynamic Graphs: This project focusses on designing Big Data platforms to support analytics and algorithms for operating over large, dynamic graphs. The emphasis will be on runtime aspects such as programming and query models for time-varying and dynamic graphs, query and execution planning, distributed scheduling and adaptivity over elasticity Cloud Virtual Machines and containers, management of distributed graph storage, etc. This systems topic will complement the above topic on analytics and algorithms for dynamic graphs. Students should have strong programming skills, preferably in Java, and be comfortable working with large code bases, Big Data platforms like Apache Hadoop and Giraph. 
  • Fog Computing for IoT Applications: Fog computing refers to the use of edge devices such as Smart Phones and Raspberry Pis, and Cloud Virtual Machines in data centers collaboratively. Internet of Things (IoT) domains is seeing the rapid rollout of edge devices with sensors ranging from smart power meters and pollution monitors, to traffic cameras. Performing analytics over such fast data streams in real-time requires that local and peer-to-peer computing on the edge complements the large-scale computing on the Cloud. Such analytics may span from complex event processing to deep learning networks. Existing Big Data platforms are not designed for such fog computing, and this project will explore topics in this regard. The outcome of this project will support the Smart Campus project being conducted at IISc. Students should have strong programming skills, preferably in Java and Python.

4. Computational Mathematics Group (Sashikumar Ganesan)

Title: Finite element modeling
The overall scope of the collaboration is to develop a customized finite element code and 3-D mesh to perform patient specific inverse finite element simulations for quantifying the non-linear hyperelastic properties of the human cornea. The following are the specific aims:

1. Customized meshing:
The human cornea is an irregular geometry that cannot be mesh with mapped settings. At the same time, the corneal collagen distribution follows an ordered pattern, which can be described with continuous functions. Since we propose to implement fiber based hyperelastic material models, a customized 3-D mesh with brick elements generator will be developed using 3-D measurements of corneal geometry.

2. Finite element solver:
The 3-D mesh will be populated with appropriate boundary conditions and material model. To solve this model, we propose to develop a parallel finite element code.

3. Inverse simulations:
General finite element simulations are forward where material properties are known. However, we need to determine the patient material properties in this case. The experimental measurement will be obtained from air-puff applanation (Corvis-ST, OCULUS Optikgerate Gmbh, Germany). Using the measurement and tools developed under (1) and (2), an inverse routine will be set up to determine the material properties. A combination of global and iterative search methods may be used so as to optimize the time required to reach a unique solution fit to the material model parameters.

This project is a collaborative work with Dr. Abhijit Sinha Roy, PhD, Chief Scientist, Imaging and Biomechanics,  Narayana Nethralaya Bangalore.
There will be a possibility for summer internship at Narayana Nethralaya Bangalore.

================================
Title: Modeling and Simulation of Energy Extraction Processes in Fuel Cells

Fuel cell is one of the promising renewable energy technologies that generates electricity by a chemical reaction. Most importantly, fuel cells   generate electricity with very little pollution, and emits a harmless byproduct, water. Unlike batteries, fuel cells   produce electricity as long as fuel (hydrogen) is supplied. Nevertheless, fuel cell is often compared to batteries as it is used as primary and backup power source.

Computational models are cost effective way  to understand and optimize the electrochemical energy extraction processes in fuel cells. The model consists of coupled nonlinear partial differential equations (PDEs), and the solution of these PDEs are very challenging.  The aim of the project is to develop an accurate and efficient numerical (finite element) scheme for simulation of electrochemical processes.  This project also involves   parallel implementation of the developed numerical scheme in our in-house finite element package ParMooN.

This project is a collaborative work with Dr. Vijay Anand Sethuraman, Ph.D.  Department of Materials Engineering
================================
Title: Computational Fluid Dynamics

Topic: Modeling and Simulation of Multiphase Flows

Free surface and two-phase flows are encountered in many applications such as spray cooling, surface coating, lab-on-a-chip, cooling in nuclear reactors, etc. The accurate numerical computation of interface flows is still a challenging task. An important issue is the precise inclusion of the surface force which compresses the surface tension and the local curvature of the free surface/interface. In addition, the presence of non-uniform distribution of surfactants on the interface induces Marangoni forces. Further, adsorption and desorption of surfactants between the interface and the bulk phase increase the complexity of the numerical computations.

Topic: Modeling and Simulation of Turbulent Flows in Time-dependent Domains

Simulations of population balance systems can be used to study the behavior of crystallization, polymerization, pharmaceutical productions, dispersed phase distribution in multiphase flows, growth of microbial and cell populations.

The aim of this project is to develop an efficient and robust numerical scheme for simulating  population balance systems with two internal coordinates in time-dependent domains.  This project also involves   implementation of the developed numerical scheme in our in-house finite element package ParMooN.

Research Areas: Crystallization Processes; Computational Mathematics; Modeling and Simulation; Finite Element Methods; Moving boundaries


5. SAFRAN lab (Sivaram Ambikasaran)


6. MARS lab (Sathish Vadhiyar)

i. Betweenness centrality on single and multi-GPU systems
Betweenness centrality is one of the important graph applications for social networks. This project aims to build efficient betweenness centrality algorithm for single and multiple GPU systems involving hybrid computations on CPUs and GPUs.
ii. Hybrid CPU-GPU programming models for Graph Applications
This projects aims to develop efficient programming models and interfaces, supported by efficient runtime stragies for CPU-GPU computations of graph applications.
iii. Fault Tolerant MPI using Adaptive process replication
This project will build a fault tolerant MPI library using process replication strategies.

7. MTech projects (non exhaustive) at the MALL lab (Partha Pratim Talukdar)

  • Building a neural conversational AI agent
  • Geolocation of tweets
  • Fast event coreference resolution and normalization
  • Fast event discovery from Twitter and other social media
  • Event extraction from social media
  • Large-scale representation learning
Strong programming background and exposure to machine learning is needed.

8. MTech projects at Video Analytics lab (R. Venkatesh Babu)
Criteria : Given the advanced nature of projects, we expect students to have a strong background in fundamentals of Image Processing and/or Pattern Recognition. There will be a formal screening interview before the student is selected.
  • Crowd counting in videos
  • Semantic scene segmentation using Attention, Recurrence and Reinforcement
  • Deep structured semantic segmentation of objects
  • Semantic segmentation of hand-drawn visual scenes

9. MTech projects at Biomolecular Computation Lab (Debnath Pal)

Title of the Project: Information Theoretical Analyses on the Clustering of Functional Protein Segments from Geometrical and Topological Properties of Peptides

It has been shown [1] that sequence entropy of protein fragments, obtained using a geometric clustering algorithm, can help analyse and identify functionally important segments, the peptides, in protein molecules. Use of Information content (IC) values of such clustered fragments using Shannon’s information measure [2] used for that purpose provides an useful direction for identifying / classifying functionally important protein segments. However, some more information theoretical analyses using other information content measures, derived from that of Shannon [2], are believed to help analyse the functional aspects of protein fragments in greater details. It is also belived that the consideration of some topological aspects of peptide structure [3] would further help in this regard.

In doing that, two more information theoretical measures will be used for this work in addition to IC used in earlier work. One of these two other measures is the Total Information Content (TIC) due to Brillouin [4] proposed to address negentropy principle in biological system. This measure would take into account the size of the clusters that contain the fragments since it is obtained by multiplying IC with the number of fragments in the cluster. This may help analyse in a different way to see whether larger clusters have any significant contribution toward identifying functionally important fragments in comparison with the clusters containing fewer number of fragments. The other information theoretical measure of interest is Complementary Information Content (CIC) that gives a measure of unavailable information in a system [5]. This will be used to investigate whether there is any significant difference in the trend of sequence entropy values from that obtained using IC earlier [1] and help analyse the functional aspects better from a complementary point of view. While the earlier work [1] was done on the basis of the gepmetrical properties of the peptides, in this work some topological properties [3] will also be considered to investigate whether it can help identify functionally important protein fragments better. In the present work, both mathematical [5] and computational studies will be carried out in addressing the proposed objective.

References:

  1. Manikandan, K., Pal, D., Ramakumar, S., Brener, N. E., Iyenger, S. S. and Sitaraman, G., Functionally Important Segments in Proteins Dissected using Gene Ontology and Geometric Clustering of Peptide Fragments, Genome Biology, 9, R52 (2008).
  2. Shannon, C. And Weaver, W., The Mathematical Theory of Communication, University of Illinois Press, Urbana, 1949.
  3. Raychaudhury, C., Banerjee, A., Bag, P. And Roy, S., Topological Shape and Size of Peptides: Identification of Potential Allele Specific Helper T Cell Antigenic Sites, J. Chem. Inf. Comput. Sci. 39, 248-254 (1999).
  4. Brillouin, L., Science and Information Theory, Academic Press, New York. 1956.
  5. Raychaudhury, C. and Pal, D., Information Content Measures and Prediction of Physical Entropy of Organic Compounds, in: Mathematical Foundations and Applications of Graph Entropy (Eds. Mathias, D. et al.), Wiley-VCH, Weinheim, Germany, 2016.