Abstract

Field Programmable Gate Arrays(FPGA) exceed the computing power of software based implementations by breaking the paradigm of sequential execution and accomplishing more per clock cycle by enabling hardware level parallelization at an architectural level. As has been proved in already published research works, introducing parallel architectures for a computationally intensive algorithm like Rapidly Exploring Random Trees(RRT) will result in an exploration that is fast, dense and uniform. FPGA based combinatorial architecture, which is one of the already published research works, delivers superlative speed-up but consumes very high power. The second of the already published research works, FPGA based hierarchical architecture, delivers relatively lower speed-up with acceptable power consumption levels. To combine the qualities of both, a hybrid architecture, that encompasses both combinatorial and hierarchical architecture, is designed. To determine the design parameters of the hybrid architecture, a cost function comprised of fundamentally inversely related speed-up and power parameters, as mentioned above, is formulated. This maximization of cost function, with its associated constraints, is then mathematically solved using a modified branch and bound, that leads to optimal deduction of the design parameters of the hybrid architecture. Via empirical experiments, it is observed that this hybrid architecture delivers the highest performance-per-watt out of the three architectures for differential, quad-copter and fixed wing kinematics, in environments of varying geometric complexity. The empirical experiments also confirmed that the hybrid architecture is scalable with 1.) Increase in the environment's geometric complexity and 2.) Increase in kinematic complexity of the robot.

Abstract

Efficient and real time segmentation of color images has a variety of importance in many fields of computer vision such as image compression, medical imaging, mapping and autonomous navigation. Being one of the most computationally expensive operation, it is usually done through software implementation using high-performance processors. In robotic systems, however, with the constrained platform dimensions and the need for portability, low power consumption and simultaneously the need for real time image segmentation, we envision hardware parallelism as the way forward to achieve higher acceleration. Field-programmable gate arrays (FPGAs) are among the best suited for this task as they provide high computing power in a small physical area. They exceed the computing speed of software based implementations by breaking the paradigm of sequential execution and accomplishing more per clock cycle operations by enabling hardware level parallelization at an architectural level. In this paper, we propose three novel architectures of a well known Efficient Graph based Image Segmentation algorithm. These proposed implementations optimizes time and power consumption when compared to software implementations. The hybrid design proposed, has notable furtherance of acceleration capabilities delivering atleast 2X speed gain over other implementations, which henceforth allows real time image segmentation that can be deployed on Mobile Robotic systems.

Abstract

Real time calculation of inverse kinematics (IK) with dynamically stable configuration is of high necessity in humanoid robots as they are highly susceptible to lose balance. This paper proposes a methodology to generate joint-space trajectories of stable configurations for solving inverse kinematics using Deep Reinforcement Learning (RL). Our approach is based on the idea of exploring the entire configuration space of the robot and learning the best possible solutions using Deep Deterministic Policy Gradient (DDPG). The proposed strategy was evaluated on the highly articulated upper body of a humanoid model with 27 degree of freedom (DoF). The trained model was able to solve inverse kinematics for the end effectors with 90% accuracy while maintaining the balance in double support phase.

Abstract

Being the ever-growing online encyclopedia, Wikipedia requires a keen investigation about which articles are to be included for it to maintain its indispensability. To prevent unnecessary articles from being included, official guidelines of Wikipedia demand these named entities to meet" notability" standards for their article inclusion. In this paper, we evaluate named entities for their notability by using reliability and entity salience features. Evaluations of our system provide evidence for the viability of our solution as an alternative to the manual decisions made by the reviewers for inclusion of an article using the notability rules.

Abstract

Networks have become ubiquitous in many real world appli-cations and to cluster similar networks is an important problem. There are various properties of graphs such as clustering coefficient (CC), den-sity, arboricity, etc. We introduce a measure, Clique Conversion Coeffi-cient (CCC), which captures the clique forming tendency of nodes in anundirected graph. CCC could either be used as a weighted average ofthe values in a vector or as the vector itself. Our experiments show thatCCC provides additional information about a graph in comparison torelated measures like CC and density. We cluster the real world graphsusing a combination of the features CCC, CC, and density and show thatwithout CCC as one of the features, graphs with similar clique formingtendencies are not clustered together. The clustering with the use ofCCC would have applications in the areas of Social Network Analysis,Protein-Protein Interaction Analysis, etc., where cliques have an impor-tant role. We perform the clustering of ego networks of the YOUTUBEnetwork using values in CCC vector as features. The quality of the clus-tering is analyzed by contrasting the frequent subgraphs in each cluster.The results highlight the utility of CCC in clustering subgraphs of alarge graph

Abstract

Time and frequency are two important dimensions to determine the interestingness of a pattern in a database. Periodic-frequent patterns are an important class of regularities that exist in a database with respect to these two dimensions. Current studies on periodic-frequent pattern mining have focused on discovering full periodic-frequent patterns, i.e., finding all frequent patterns that have exhibited complete cyclic repetitions in a database. However, partial periodic-frequent patterns are more common due to the imperfect nature of real-world. This paper proposes a flexible and generic model to find partial periodic-frequent patterns. A new interesting measure, periodic-ratio, has been introduced to determine the periodic interestingness of a frequent pattern by taking into account its proportion of cyclic repetitions in a database. The proposed patterns do not satisfy the anti-monotonic property. A novel pruning technique has been introduced to reduce the search space effectively. A pattern-growth algorithm to find all partial periodic-frequent patterns has also been presented in this paper. Experimental results demonstrate that the proposed model can discover useful information, and the algorithm is efficient.

Abstract

We propose an interactive visual analytics system for exploring spatio-temporal data in VAST 2017 Mini-Challenge-1. As part of this challenge, we are expected to determine repeating, seasonal and unusual cars' movements in Lekagul park dataset. We use varied visualizations such as heatmap, sequential sunburst and line plots. Further, we have developed a web deployable ad-hoc system for displaying spatio-temporal information on geographical map

Abstract

A temporal database is a collection of transactions, ordered by their timestamps. Discovering periodic patterns in temporal databases has numerous applications. However, to the best of our knowledge, no work has considered mining periodic patterns in temporal databases where items have dissimilar support and periodicity, despite that this type of data is very common in real-life. Discovering periodic patterns in such non-uniform temporal databases is challenging. It requires defining (i) an appropriate measure to assess the periodic interestingness of patterns, and (ii) a method to efficiently find all periodic patterns. While a pattern-growth approach can be employed for the second sub-task, the first sub-task has to the best of our knowledge not been addressed. Moreover, how these two tasks are combined has significant implications. In this paper, we address this challenge. We introduce a model to assess the periodic interestingness of patterns in databases having a non-uniform item distribution, which considers that periodic patterns may have different period and minimum number of cyclic repetitions. Moreover, the paper introduces a pattern-growth algorithm to efficiently discover all periodic patterns. Experimental results demonstrate that the proposed algorithm is efficient and the proposed model may be utilized to find prior knowledge about event keywords and their associations in Twitter data.

Abstract

Community-driven Knowledge Sharing (KS) plat-forms have gained immense popularity in recent years among the Internet users to seek, learn & share information and expertise.These platforms encourage rich content by recognizing users’ contributions; measured as the reputation on platforms. Thus,peers on the platform dissuade other users to post low quality content by closing, disliking or not answering their questions.The aggressive attitude of the community leads to a negative impact on the users’ experience (especially for newbies) and the users tend to lose interest in the platform. Since the quality management on KS platforms is a necessity, we aim to emphasize the need of a mechanism to help users in improving their questions. To study this, we perform experiments on the Stack Overflow (SO) platform, however, the study can be adapted for other KS platforms.The SO community aggressively mark questions, even with slightest deformity, as closed. The non-triviality of reopening process and the rigorous review mechanism can be handled only by proper editing of the closed questions. Thus we present first-of-a-kind study, to assist the SO platform users, in the reopening process of their closed questions. We build predictive models to suggest the users that if their edited version of closed question will lead to a successful reopen or not. This can assist users at large by retracting them from entering the review process with improper edits. To learn these models effectively, we consider the user categories on the platform, based on their reputation: established& non-established. In addition to being the major contributor to closed questions, the non-established users have lower odds of getting their closed questions reopened than established users.Thus, we leverage the better editing skills of established users to learn question edit models by employing their reopened closed questions. We present the results of predictive model and suggest insights on the useful edits in the reopening process on the SO platform

Abstract

Periodic Frequent patterns (PFPs) are an important class of regularities that exist in a transactional database. In the literature,pattern growth-based approaches to mine PFPs have be proposed by considering a single machine. In this paper, we propose a Map-Reduce framework to mine PFPs by considering multiple machines. We have proposed a parallel algorithm by including the step of distributing transactional identifiers among the machines. Further, the notion of partition summary has been proposed to reduce the amount of data shuffled among the machines. Experiments on Apache Spark’s distributed environment show that the proposed approach speeds up with the increase in number of machines and the notion of partition summary significantly reduces the amount of data shuffled among the machines.