Department of Computer Science and Engineering

Indian Institute of Technology Hyderabad (IITH) is a premier institute of science and technology established in 2008. IITH has been consistently ranked in the&nbsp;<a href="https://iith.ac.in/reports/" rel="noopener noreferrer">top 10 institutes</a>&nbsp;in India for Engineering&nbsp;<a href="https://www.nirfindia.org/2021/EngineeringRanking.html" rel="noopener noreferrer" target="_blank">according to NIRF</a>&nbsp;making it one of the most coveted schools for science and technology in the country.

IITH offers&nbsp;<a href="https://iith.ac.in/academics/under-graduate/">undergraduate programs</a>&nbsp;in all the classical engineering disciplines, applied sciences, design, as well as several modern interdisciplinary areas. Students are given a flexibility to explore a broad set of areas, and potentially pursue a minor or double major in a discipline that is not their own. Students who wish to seek a deeper understanding of their own discipline are strongly encouraged to get involved in cutting-edge research with the help of a faculty to mentor them, and earn an Honors in their own field.

The very foundation of IIT Hyderabad is based on&nbsp;<a href="https://iith.ac.in/research/">research and innovation</a>. The vibrant research culture is evident from the number of patents and publications that IITH produces consistently every year. IITH offers&nbsp;<a href="https://iith.ac.in/academics/programmes-offered/">graduate programs</a>&nbsp;at both a masters, and a doctoral level, in several diverse areas. There are separate programs for technology, design, science, and liberal arts. The&nbsp;<a href="https://iith.ac.in/academics/post-graduate/">MTech program</a>&nbsp;is offered in various modes and durations to cater to the ever-growing need of postgraduate level professionals in the country.

IITH encourages and supports&nbsp;<a href="https://iic-iith.netlify.app/index.html" rel="noopener noreferrer" target="_blank">innovation and entrepreneurship</a>&nbsp;at every stage. The minor program in entrepreneurship is open to all students of IITH, and many of the courses are offered by industrialists who are entrepreneurs themselves. There are various&nbsp;<a href="https://iith.ac.in/research/centres-incubators/">innovation centers and incubators</a>&nbsp;that promote entrepreneurship and help mentor young innovators. IITH has been very successful in building tie-ups with leading academic institutions around the globe.

IITH enjoys a very special relationship with Japanese universities and industries that goes beyond academic and research collaborations. In fact, some of the iconic buildings in IITH campus will carry the signature of Japanese architecture.

IITH creates a unique holistic ecosystem for education that offers interactive learning, a very flexible academic structure, cutting-edge research, strong industry collaboration, and entrepreneurship. This is an environment which enables students and faculty to translate their dreams into realities.

IIT Hyderabad

Purpose: The innovation in big data is increasing day by day in such a way that the conventional software tools face several problems in managing the big data. Moreover, the occurrence of the imbalance data in the massive data sets is a major constraint to the research industry. Design/methodology/approach: The purpose of the paper is to introduce a big data classification technique using the MapReduce framework based on an optimization algorithm. The big data classification is enabled using the MapReduce framework, which utilizes the proposed optimization algorithm, named chicken-based bacterial foraging (CBF) algorithm. The proposed algorithm is generated by integrating the bacterial foraging optimization (BFO) algorithm with the cat swarm optimization (CSO) algorithm. The proposed model executes the process in two stages, namely, training and testing phases. In the training phase, the big data that is produced from different distributed sources is subjected to parallel processing using the mappers in the mapper phase, which perform the preprocessing and feature selection based on the proposed CBF algorithm. The preprocessing step eliminates the redundant and inconsistent data, whereas the feature section step is done on the preprocessed data for extracting the significant features from the data, to provide improved classification accuracy. The selected features are fed into the reducer for data classification using the deep belief network (DBN) classifier, which is trained using the proposed CBF algorithm such that the data are classified into various classes, and finally, at the end of the training process, the individual reducers present the trained models. Thus, the incremental data are handled effectively based on the training model in the training phase. In the testing phase, the incremental data are taken and split into different subsets and fed into the different mappers for the classification. Each mapper contains a trained model which is obtained from the training phase. The trained model is utilized for classifying the incremental data. After classification, the output obtained from each mapper is fused and fed into the reducer for the classification. Findings: The maximum accuracy and Jaccard coefficient are obtained using the epileptic seizure recognition database. The proposed CBF-DBN produces a maximal accuracy value of 91.129%, whereas the accuracy values of the existing neural network (NN), DBN, naive Bayes classifier-term frequency–inverse document frequency (NBC-TFIDF) are 82.894%, 86.184% and 86.512%, respectively. The Jaccard coefficient of the proposed CBF-DBN produces a maximal Jaccard coefficient value of 88.928%, whereas the Jaccard coefficient values of the existing NN, DBN, NBC-TFIDF are 75.891%, 79.850% and 81.103%, respectively. Originality/value: In this paper, a big data classification method is proposed for categorizing massive data sets for meeting the constraints of huge data. The big data classification is performed on the MapReduce framework based on training and testing phases in such a way that the data are handled in parallel at the same time. In the training phase, the big data is obtained and partitioned into different subsets of data and fed into the mapper. In the mapper, the features extraction step is performed for extracting the significant features. The obtained features are subjected to the reducers for classifying the data using the obtained features. The DBN classifier is utilized for the classification wherein the DBN is trained using the proposed CBF algorithm. The trained model is obtained as an output after the classification. In the testing phase, the incremental data are considered for the classification. New data are first split into subsets and fed into the mapper for classification. The trained models obtained from the training phase are used for the classification. The classified results from each mapper are fused and fed into the reducer for the classification of big data. © 2020, Emerald Publishing Limited.

World University root node

Data Technologies and Applications

Chicken swarm foraging algorithm for big data classification using the deep belief network classifier

Cognitive science has shown results on how human subjects (human children) and other mammals adapt to an entirely novel task (depth measurement) by understanding the association with already learned tasks (self-motion, shoulder movements) without receiving an explicit supervision. Motivated by such prior work, in this chapter, we present a meta-regression algorithm that regresses model parameters of zero-shot tasks from the model parameters of known tasks and the correlation of known and zero-shot tasks. The proposed method is evaluated on the Taskonomy dataset [54] considering surface normal estimation, depth estimation, room layout and camera pose estimation as zero-shot tasks. Our proposed methodology outperforms state-of-the-art models (which use ground truth) on each of our zero-shot tasks, showing promise on zero-shot task transfer. We also conducted extensive experiments to study the various choices of our methodology, as well as showed how the proposed method can also be used in transfer learning. To the best of our knowledge, this is the first such effort on zero-shot learning in the task space. © Springer Nature Switzerland AG 2020.

Domain Adaptation in Computer Vision with Deep Learning

Zero-Shot Task Transfer

Given a bipartite graph G=(U⊎V,E), a linear representation of the transversal matroid associated with G on the ground set U, can be constructed in randomized polynomial time. In fact one can get a linear representation deterministically in time 2O(m2n), where m=|U| and n=|V|, by looping through all the choices made in the randomized algorithm. Other important matroids for which one can obtain linear representation deterministically in time similar to the one for transversal matroids include gammoids and strict gammoids. Strict gammoids are duals of transversal matroids and gammoids are restrictions of strict gammoids. We give faster deterministic algorithms to construct linear representations of transversal matroids, gammoids and strict gammoids. All our algorithms run in time (mr)mO(1), where m is the cardinality of the ground set and r is the rank of the matroid. In the language of parameterized complexity, we give an XP algorithm for finding linear representations of transversal matroids, gammoids and strict gammoids parameterized by the rank of the given matroid. © 2018 Elsevier B.V.

Theoretical Computer Science

Linear representation of transversal matroids and gammoids parameterized by rank

We study the question of finding a set of at most k edges, whose removal makes the input n-vertex graph a disjoint union of s-clubs (graphs of diameter s). Komusiewicz and Uhlmann [DAM 2012] showed that CLUSTER EDGE DELETION (i.e., for the case of 1-clubs (cliques)), cannot be solved in time 2o(k)nO(1) unless the Exponential Time Hypothesis (ETH) fails. But, Fomin et al. [JCSS 2014] showed that if the number of cliques in the output graph is restricted to d, then the problem (d-CLUSTER EDGE DELETION) can be solved in time O(2O(dk)+m+n). We show that assuming ETH, there is no algorithm solving 2-CLUB CLUSTER EDGE DELETION in time 2o(k)nO(1). Further, we show that the same lower bound holds in the case of s-CLUB d-CLUSTER EDGE DELETION for any s≥2 and d≥2. © 2020 Elsevier Inc.

Journal of Computer and System Sciences

Subexponential algorithm for d-cluster edge deletion: Exception or rule?

Timely forecast of traffic is very much needed for smart cities, which allows travelers and government agencies to make various decisions based on traffic flow. This will result in reduced traffic congestion and carbon dioxide emission. However, traffic forecasting is a challenging task due to the highly complex traffic pattern. Standard time series techniques may not be able to capture the nonlinear and noisy nature of the traffic flow. In this paper, we investigate how the deep learning models capture these characteristics and provide better predictive performance over standard time series and regression models. We compare the performances of state-of-the-art deep learning models on two traffic flow data sets and show their effectiveness in traffic flow prediction over traditional models. © 2020 IEEE.

2020 IEEE Region 10 Symposium, TENSYMP 2020

Traffic Forecasting with Deep Learning

We propose IR2VEC, a Concise and Scalable encoding infrastructure to represent programs as a distributed embedding in continuous space. This distributed embedding is obtained by combining representation learning methods with flow information to capture the syntax as well as the semantics of the input programs. As our infrastructure is based on the Intermediate Representation (IR) of the source code, obtained embeddings are both language and machine independent. The entities of the IR are modeled as relationships, and their representations are learned to form a seed embedding vocabulary. Using this infrastructure, we propose two incremental encodings: Symbolic and Flow-Aware. Symbolic encodings are obtained from the seed embedding vocabulary, and Flow-Aware encodings are obtained by augmenting the Symbolic encodings with the flow information. We show the effectiveness of our methodology on two optimization tasks (Heterogeneous device mapping and Thread coarsening). Our way of representing the programs enables us to use non-sequential models resulting in orders of magnitude of faster training time. Both the encodings generated by IR2VEC outperform the existing methods in both the tasks, even while using simple machine learning models. In particular, our results improve or match the state-of-the-art speedup in 11/14 benchmark-suites in the device mapping task across two platforms and 53/68 benchmarks in the thread coarsening task across four different platforms. When compared to the other methods, our embeddings are more scalable, is non-data-hungry, and has better Out-Of-Vocabulary (OOV) characteristics. © 2020 ACM.

ACM Transactions on Architecture and Code Optimization

IR2Vec: LLVM IR Based Scalable Program Embeddings

Multimedia data is known for its variety and also for the difficulty that comes in extracting relevant features from multimedia data. Owing to which the collaborative recommendation systems have found their foothold in multimedia recommender systems. However, modern-day multimedia sites have tons of user history in the form of user feedback, reviews, votes, comments, and etc. We can use these social interactions to extract useful content features, which can then be used in content based recommendation system. In this paper, we propose a novel hybrid recommender system that combines the content and collaborative systems using a Bayesian model. We substitute the concrete textual content with a sparse tag information. Extensive experiments on real-world dataset show that tags significantly improves the recommendation performance for multimedia data. © 2020 IEEE.

Proceedings - 2020 IEEE 6th International Conference on Multimedia Big Data, BigMM 2020

Tag Boosted Hybrid Recommendations for Multimedia Data

Indirect taxation is a significant source of livelihood for any nation. Tax evasion inhibits the economic growth of a nation. It creates a substantial loss of much needed public revenue. We design a method to single out taxpayers who evade indirect tax by dodging their tax returns. Towards this, we derive six correlation parameters (features), three ratio parameters from tax return statements submitted by taxpayers, and another parameter based on the business interactions among taxpayers using the TrustRank algorithm. Then we perform spectral clustering on taxpayers using these ten parameters (features). We identify taxpayers located at the boundary of each cluster by using kernel density estimation, which are further investigated to single out tax evaders. We applied our method on the iron and steel taxpayer’s data set provided by the Commercial Taxes Department, Government of Telangana, India. © Springer Nature Switzerland AG 2020.

Lecture Notes in Business Information Processing

Detecting tax evaders using trustrank and spectral clustering

5G Mobile network will reap the benefits from key technologies like Software Defined Networking and Network Function Virtualization. Cloud Radio Access Network architecture (Cloud RAN) is proven to be a promising architecture, but fully centralized Cloud RAN imposes a great bandwidth requirement in the fronthaul link. Different functional split options for 5G RAN have been proposed which lead to a trade-off between centralization and bandwidth requirement. Functional split at different granularity such as per cell, per logical network (slice), per user, or per bearer, have been an area of interest. To explore the effect of slice granularity in adaptive splits for slices, we formulate slice centric functional split in 5G RAN as an ILP to maximize centralization of baseband processing. By varying the slice granularity from macro slicing to micro slicing, we observe how slice centric split can impact centralization benefit of the network. We show that with increasing slice granularity slice centric split can render more centralization benefit in some scenarios but a trade off exists between centralization benefit and migration cost in the network which should be carefully considered in real deployment scenario. © 2020 IEEE.

Proceedings of the 2020 IEEE Conference on Network Softwarization: Bridging the Gap Between AI and Network Softwarization, NetSoft 2020

Impact of slice granularity in centralization benefit of 5G radio access network

Network slicing is an integral part of 5G, which supports next-generation wireless applications over a shared network infrastructure. It paves the way to leverage the full potential of 5G by increasing the efficiencies through differentiation and faster time-to-market. In this work, we propose a Mobile Virtual Network Operator (MVNO) Slice Resource Allocation Architecture (MSRAA) for supporting different network slices in the 5G data plane. MSRAA supports QoS parameters, including Guaranteed Bit Rate (GBR) and Maximum Delay Budget. Using long short-term memory (LSTM) neural networks, we predict network slices bandwidth requirements for efficiently allocating the resources. To reduce revenue loss to the network operators due to forecasting errors, the proposed Bandwidth Admission Control (BAC) algorithm, reallocates resources from lower priority slices (e.g., best-effort users) to higher priority slices (e.g., guaranteed service users). Using Mondrain Random Forests in our Delay Admission Control (DAC) algorithm, we predict the end-to-end delay and admit flows into slices that can satisfy delay requirements. We implement MSRAA on our advanced 5G Core testbed and evaluate User Service Request (USR) acceptances and do a complete cost-benefit analysis of our architecture. We show that for eMBB-GBR and eMBB-Non-GBR slices, our algorithm is showing a significant reduction in costs. © 2020 IEEE.