Research Areas: Mobile Wireless Networks, 5G Networks and Systems, Cloud RAN, Edge Cloud.
With the increasing number of connected devices (5.6 billion smart phone subscriptions expected at end of 2019), there is an exponential growth in data usage by smart devices such as smart phones and tablets (10 X growth in mobile data traffic between 2013 and 2019). The fifth generation (5G) cellular technology would be the beginning of the realization of Internet of Things (IoT) where everything would be connected to the Internet. This new generation essentially redefine the way these devices communicate with each other. The main key requirements of this new generation mobile system are Gpbs data rate anywhere in the cell and less than a msec end-to-end latency. Also as we move on to this new 5G era, more emphasis would be given for the Quality of Experience (QoE) of next generation mobile services such as UHD video streaming, virtual reality, tactile Internet, autonomous driving, etc. These requirements are very hard to achieve with the existing cellular systems. Different technological advancements are needed to achieve these stringent requirements of the next generation mobile services. We work on various key research issues that need to be addressed to realize the success of next generation mobile technology such as inter-working of multiple radio access technologies (convergence of 4G, 5G, and WiFi), converged cloud radio access network, edge content distribution, low latency transport, etc. We also develop test-beds and prototypes to demonstrate the working of our developed technologies in a real-world scenario. For more info, visit http://newslab.iith.ac.in/newslab/
Research Areas: Converged Radio Access Networks , SDN, IoT and Green ICT
The proliferation and penetration of smart phones and IoT devices is expected to have profound impact on world economy. In order to support billions of wirelessly connected devices, and tackle exponentially increasing traffic demands and ever-increasing QoE and energy efficiency requirements of diverse services and applications being carried over communication network infrastructure, a heterogeneous converged network (HCN) architecture integrating heterogeneous radio access technologies including various generations of cellular (e.g., 4G/5G), Wi-Fi (e.g., IEEE 802.11n/ac), Bluetooth and ZigBee, and wireless core network technologies is required. We expect that HCN contains ultra-dense deployment of network elements (e.g., base stations and access points) to boost network coverage and capacity. My team is carrying out R&D work on various research issues like novel (cloud) convergence architectures, seamless mobility, load-balancing, interference management, co-existence protocols, and optimal resource allocation to realize energy-efficient and spectrally-efficient HCNs. It is promising to design and develop HCNs on open-hardware platforms by exploiting recent advances in network functions virtualization (NFV), virtualization techniques of cloud computing and software-define networking (SDN) which direct flexible, dynamically configurable network elements to provide on-demand customized services to traffic demands which may be dynamic in time and space. We build testbeds and do prototyping to assess benefits of technology pilots in a real-world setting. For more info, visit http://newslab.iith.ac.in/newslab/
Research Areas : Video Content Analysis, Machine Learning,Sparsity-Based Methods, Deep Learning
With the advent of relatively inexpensive video recording mediums there is a bombardment of videos on the Internet. However, most of it is uncategorized and hence the need for video content analysis. The main objective of our research lab VIGIL (VIsual learning and Intelligence group) is to address various issues in feature extraction, video segmentation, video classification, and video abstraction for exploring visual big data. Some of the important studies include: (a) exploring features to describe video content, (b) video segmentation into meaningful units (shots/scenes/stories), (c) video genre classification, (d) exploring compressive sensing, deep learning and kernel methods for various tasks in video content analysis like video classification, clustering, dimension reduction, event detection and activity recognition, (e) expression recognition in videos, (f) activity classification in videos with cloud based and distributed architectures. Other research areas include content based medical image classification and retrieval using sparse representation and dictionary learning techniques and also de-duplication of biometric traits for large-scale applications. More details can be obtained at http://www.iith.ac.in/vilg/
Research Areas : Theoretical computer science, computational complexity.
My research interests lie primarily in Computational Complexity Theory. The field of computational complexity is a formal study of the difficulty (complexity) involved in computing various functions. Such a study involves a notion of reductions using which one can show that a certain function is harder to compute than another. Such reductions have given rise to a rich classification of functions into 'complexity classes'. One of the primary goals of the field is to understand these complexity classes and the relationships between them. One famous problem in this context is the P vs NP problem.
I am particularly interested in studying computation under very strong restrictions. I typically work on problems that involve studying circuits and circuit complexity classes that are very restricted in terms of resources.
Research Areas: Blockchain, Internet Architecture, Software-Defined Networking (SDN), Network Functions Virtualization (NFV), Network Operation, Post-Disaster Networking, Any Fun Applications
Network is fundamental for everything to allow it be connected and collaborating with something else. In addition, things need trust and security so that such a collaboration can become more useful and valuable. Exploring both of practical applications and fundamental platforms and underlying technologies, my research interest covers various topics on Blockchain, Internet architectures, these applications, and collaboration with something that we don't know yet!!
Research Areas : Databases, Data Mining, HCI, Information Retrieval, Information Visualization
During the last few decades, database researchers have greatly improved the capability of databases both in terms of performance and functionality. But users’ limited knowledge of what is contained in various databases and lack of technical expertise makes it very hard for them to directly interact with databases in a meaningful way. Managing a database system thus usually requires an army of database administrators, consultants, and other technical experts all busily helping users get data in and out of the database. To make databases more accessible to users, many researchers in the past few years have started to look at the usability aspect of databases, such as usability in data storage, data access, data presentation, etc. I use a combination of machine learning, data mining, statistics, and information visualization to help users more effectively explore through huge volumes of relational data. Apart from databases, I also work in Big Data management and analysis. Although there are many existing Big Data platforms, they are not very usable because it requires significant amount of technical expertise to use those platforms for each application. My team is working on designing a top-layer software system that will make the existing Big Data platforms more user-friendly and easily accessible. The system will primarily facilitate understanding and interpretation of the result returned by the existing Big Data platforms. The system is being designed using high performance computing techniques to make the system have both high usability and performance.
Research Areas: Applied Algebraic Topology, Topological Data Analysis (TDA), Machine Learning, Geometry
Research Areas: Computational Algebra, Symbolic Computation, Lattice Cryptography
My research interests lie broadly in theoretical computer science, particularly in areas that have deep connections with mathematics such as computational algebra, symbolic computation, cryptography and mathematical software.
Several real world problems can be modeled as non-linear polynomial equations and determining their solutions is an important and challenging problem. My research focuses on developing efficient algorithmic tools to compute the algebraic properties of polynomial equations where the coefficients are from rings such as the ring of integers. Specifically, I look at extending Groebner basis techniques, a standard algorithmic tool for multivariate polynomial rings over fields, to polynomials over rings.
Polynomial rings over rings have applications in cryptography, coding theory, algebraic geometry, etc. I am also interested in how these Groebner basis tools for polynomial rings over integers can be used to design different cryptographic primitives in lattice-based cryptography, an important area in post-quantum cryptography.
Research Areas: Information Retrieval, Recommender Systems, Data Mining, Machine Learning
Today, we are living in the age of data explosion. There is huge amount of data that is being stored digitally in isolated computing devices, intranets and internet. Efficient modeling, retrieval and analysis of this data is a big challenge. My work deals with understanding and addressing some of these challenges. Given below are some example (broad) research directions that I am currently pursuing or interested in working on.
- Suppose there is a data repository of some form (e.g. document corpus, social network graph, microblogs collection etc.), and a user has an information need. How do we address that information need by finding relevant results from the collection? My work in this area is targeted towards understanding of domain specific feature identification, designing ranking algorithms, fusing available data from external sources for context identification etc.
- Consider a repository containing overwhelming amount of information about a specific topic of interest (e.g. movie information, ECommerce catalog on electronic gadgets, music listings in online channel, news portal containing articles from multiple sources etc.). User is interested in that specific topic (e.g. movie, product, music, news etc.) but is not willing to or finds it difficult to specify her information need. In this scenario, how to recommend items to the user that she may like? In this theme, we are working on identifying and/or evaluating modes of user feedback, incorporating novel feedback mechanisms in existing frameworks of personalized recommendation, diversification in recommendations, understanding temporal aspects in recommendation, cross-domain recommendations etc.
- In today's web there are lots of entities (people or organizations) that generate various types of content (e.g. microblogs, articles, images, educational materials etc.), and there are entities that consume those contents. I am interested in modeling and analyzing the behaviours of content generators as well as content consumers in the space of user generated contents.
There are lots of "digitally consumable items/contents" (music, movie, image, microblogs etc.) on the web today. Also, users can watch/view these items in real world and come back to digital media to share their thoughts regarding the items. These items have different dynamics in their appeal to the users. Understanding, predicting and making use of this item dynamics in various services is another research direction that I am interested in.
Research Areas : Algorithms, Graph Theory,Combinatorics
In practice, many natural algorithmic problems turn out to be NP-hard, which means that designing efficient algorithms for them is a challenge. One approach to this is the design of parameterized algorithms, which is one of my research interests. My other research interests include all kinds of graph theoretic problems, especially graph coloring.
Head of Department & Associate Professor
Research Areas : Theoretical Computer Science: Computational Complexity, Quantum Computing, Formal Methods and their Applications
In Theoretical Computer Science, I am particularly interested in computational complexity, randomized algorithms and quantum computing. I have carried out research in quantum automata theory, algorithms and quantum cryptography in the past. I am also interested in formal methods, particularly model checking for hybrid and reactive systems and verification of security protocols. In these areas, my research group is also working on applications in addition to the theoretical aspects.
Research Areas : Programming Languages, Multi-core and GPU compilation, Static analysis, Verification and Abstract Interpretation.
The modern day architectures with multi-core and GPUs processors throw a difficult challenge to the compiler community. Compiling applications from many domains (like image processing, machine learning and big-data) so that they can run efficiently on these architectures efficiently exploiting the available resources demands advanced techniques. The polyhedral compilation community has developed a formal framework so that loop programs could be brought into a common intermediate representation and transformed so as to cater to the needs of these architectures. Polyhedral compilation broadly uses Linear and Integer Linear programming techniques to transform the input for-loop programs. It has been quite successful in these effort: open source compilers (like LLVM) and industry compilers (like the one by IBM) have incorporated these techniques. However, more needs to be done so that the range of the input programs as well as the scalability of the techniques themselves could be improved. Broadly, I am interested in any compiler research in the open-source LLVM compiler, and so run a local IITH LLVM Group. I am also interested in a wide variety of research areas related to program analysis like Domain Specific Programming Languages for Parallelization, Abstract Interpretation for verification of programs, abstract domains for scalability as well as combinatorial optimization. For more information visit : https://github.com/IITH-Compilers/
Research Areas : Machine Learning
My research interest primarily is in the area of Machine Learning, with focus on kernel methods. The main theme in most of my works is developing novel optimization formulations and algorithms, along with relevant theoretical gaurantees, that can be applied to diverse machine learning applications. The key tools that I leverage in the context of learning are Optimization and Statistics. I am currently pursuing projects on i) Kernel Embeddings ii) Causality in ML iii) Some text/vision related applications.
Research Areas : Parallel Programming, Software Transactional Memory, Distributed Systems, Algorithm Analysis .
Over the last few decades, much of the gain in software performance can be attributed to increases in CPU clock frequencies. However around 2004, 50 years of exponential improvement in the performance of sequential computers ended. Industry’s response to these changes was to introduce single-chip, parallel computers, variously known as “chip multiprocessors,” “multicore,” or “manycore” computers. In order to get a continued speedup on these processors, applications need to be able to harness the parallelism of the underlying hardware. This is commonly achieved using multi-threading. Therefore, there is an increasing need for programmers to produce applications in which multiple processes execute in parallel and coordinate to achieve some shared task. Yet writing correct and scalable multi-threaded programs is far from trivial. While it is well known that shared resources must be protected from concurrent accesses to avoid data corruption, guarding individual resources is often not sufficient. Sets of semantically related actions may need to execute atomically to avoid semantic inconsistencies. Dr. Sathya Peri currently, works on developing efficient software for managing concurrency in multi-threaded programs. Apart from this, he works on the other areas of parallel and distributed systems.
Research Areas : Software Verification, Concurrency, Constraint solving, Program Analysis, Formal Methods.
Software has penetrated almost all the aspects of our lives. Failure of a software in a critical system can not only lead to huge amount of monetary loss but it can also lead to loss of human life. Therefore, it becomes of paramount importance that software systems do not fail. My primary area of research is formal software verification. In particular, I look at verifying software that are concurrent in nature. My goal is to develop tools and techniques for software verification that can be easily used to verify real world programs. Recently, I have also started exploring reuse of software verification techniques for hardware domain. As many verifiers rely on constraint solvers, I have started efforts on developing novel encoding techniques that would make constraint solvers more efficient.
Research Areas : Big Data Analytics, Applied Algorithms
The goal of big data analytics is to help companies and governments to make more informed decisions by enabling data scientists, predictive modelers and other analytics professionals to analyze large volumes of structured and unstructured data that may be untapped by conventional business intelligence programs. We are developing algorithms and models for analyzing traffic data for providing better decision support for city police and increasing the productivity of citizens.
Research Areas: Processor architectures for AI, Computer architecture (CPUs and GPUs), machine learning on edge devices, approximate computing, autonomous driving
I work in the intersection of computer architecture and machine learning. I am interested in designing accelerators for machine learning especially neural networks, processor architectures for deep learning. Further, I am interested in accelerating machine learning tasks, especially autonomous driving using local processing (i.e., without the support of cloud). I have worked on optimizing the architecture of both CPUs and GPUs. I have proposed techniques to improve energy efficiency, performance and reliability of memory system. I have also developed open-source tools for modeling emerging 3D non-volatile memories.
Research Areas : Bayesian data analysis, probabilistic machine learning, Bayesian non-parametrics, survival analysis, text analytics.
My research interest lies in developing probabilistic machine learning and Bayesian data analysis techniques to solve real world learning problems. I have developed techniques based on probabilistic methods such as Gaussian processes, Dirichlet processes and point processes, and kernel methods to solve problems in natural language processing, information retrieval and social networks.
Parametric and non-parametric Bayesian models allow the incorporation of prior information and domain knowledge. Non-parametric Bayesian models additionally allow one to learn rich and flexible models due to their non-parametric nature and allow the model complexity to be determined by the data. This helps to overcome the problem of model selection to a great extent. I am working on developing scalable non-parametric Bayesian models and efficient inference algorithms for Big data scenario. I am also interested in event history analysis and in analysing temporal textual data. Current research applies Bayesian statistics and probabilistic models to diverse problem domains such as optimization, numerical methods, web, social networks, and healthcare.
Research Areas : Theoretical Computer Science, Randomized Algorithms, Complexity Theory, Combinatorics
The area of theoretical computer science studies computational problems and attempts to provide theoretical guarantees for easiness/hardness of these problems. I am interested in almost all aspects of theoretical computer science. One of the specific areas that I have been looking at is the area of property testing. The model in property testing assumes a typical “big data” scenario – the input is too large to be inspected by the algorithm. The size of the input prevents us from taking a pass at the entire input, it will simply take too much time. The area of property testing attempts to study those algorithms which would look at only a small part of the input, and yet make a correct answer whether the input has the desired property or not. The area of sublinear algorithms is a closely related area. Property testing algorithms came into the spot light during the last two decades and this area has seen tremendous growth in the last decade. Being a relatively new area, I feel that this area has more questions than answers and so there is a lot of scope for exciting research.
Research Areas: Machine Learning, Deep Learning, Computer Vision, Optimization
My research interests lie at the intersection of the theory and application of machine learning. With a strong interest in the mathematical fundamentals and a passion for real-world application, my group’s research has focused on being at the forefront of the field, by publishing in the best of venues in the areas of machine learning and computer vision, while at the same time, being guided by application contexts derived from real-world use. In recent years, our work on the algorithmic side of machine learning has resulted in peer-reviewed articles in non-convex optimization, explainable machine learning and deep generative models; while our applied work has resulted in application of machine learning to domains including agriculture, e-learning, security/surveillance and human behavior understanding. As machine learning and Artificial Intelligence continue to make inroads into the daily life of every human, my future research plans focus on the development of machine learning methods for newer real-world application contexts and challenges, with a strong emphasis on theoretical analysis of the proposed methods in terms of convergence, performance guarantees and generalization bounds. Please see http://www.iith.ac.in/~vineethnb/ for more details.