Computer Science and Engineering

Video question answering is the task of automatically answering questions about videos. Among query types which include identification, localization, and counting, the most challenging questions enquire about relationships among different entities. Answering such questions, and many others, require modeling relationships between entities in the spatial domain and evolution of those relationships in the temporal domain. We argue that current approaches have limited capacity to model such long-range spatial and temporal dependencies. To address these challenges, we present a novel spatio-temporal reasoning neural module which enables modeling complex multi-entity relationships in space and long-term ordered dependencies in time. We evaluate our module on two benchmark datasets which require spatio-temporal reasoning: TGIF-QA and SVQA. We achieve state-of-the-art performance on both datasets. More significantly, we achieve substantial improvements on some of the most challenging questi...

This paper presents the machine learning based machine translation system for Hindi to English, which learns the semantically correct corpus. The quantum neural based pattern recognizer is used to recognize and learn the pattern of corpus, using the information of part of speech of individual word in the corpus, like a human. The system performs the machine translation using its knowledge gained during the learning by inputting the pair of sentences of Devnagri-Hindi and English. To analyze the effectiveness of the proposed approach, 2600 sentences have been evaluated during simulation and evaluation. The accuracy achieved on BLEU score is 0.7502, on NIST score is 6.5773, on ROUGE-L score is 0.9233, and on METEOR score is 0.5456, which is significantly higher in comparison with Google Translation and Bing Translation for Hindi to English Machine Translation.

This paper presents the machine translation system (MTS) which is based on the concept of self learning of semantically correct corpus using pattern recognition. The self learning process using pattern recognition is based on Quantum Neural Network (QNN). This is a novel and new approach to recognize and learn the corpus pattern using QNN. The paper 9ppresents systematically structure of the system, machine translation system and performance results. Present procedure performs the task of translation using its knowledge gained during learning by inputting pair of sentences from source to target language. Like a person, the system also acquires the necessary knowledge required for translation in implicit form from inputting pair sentences. The performance is also compared with other ANN approaches. It has also been shown that QNN requires less training time than the traditional ANN based training. http://www.ripublication.com/irph/ijict_spl/11_ijictv3n7spl.pdf

Soft computing techniques are a good solution for the face detection. Neural network is one of the soft computing techniques, which are generally used for learning and training process. Face detection is one of the challenging problems in the image processing. The basic aim of face detection is determine if there is any face in an image. And then locate position of a face in image. Human face detected in an image can represent the presence of a human in a place. Evidently, face detection is the first step towards creating an automated system, which may involve other face processing. A novel face detection system is presented in this research work. The approach relies on neural networks, which can be used to detect faces by using FFT. The neural network is created and trained with training set of faces and non-faces. The network used is a two layer feed-forward neural network. There are two modifications for the classical use of neural networks in face detection. First, the neural network tests only the face candidate regions for faces, thus the search space is reduced. Second, the window size used by the neural network in scanning the input image is adaptive and depends on the size of the face candidate region. The objective of this work was to implement a classifier based on MLP (Multi-layer Perception) neural networks for face detection. The MLP was used to classify face and non-face patterns.

The goal of this thesis is to investigate the current techniques for fingerprint identification system. This target can be mainly decomposed into image preprocessing, feature extraction and feature match. For each sub-task, some classical and up-to-date methods in literatures are analyzed. Based on the analysis, an integrated solution for fingerprint identification system is developed for demonstration.

Handwriting recognition has attracted many researchers across the world. Recognition of online handwritten Hindi numerals is a goal of many research efforts in the pattern recognition field. This paper presents an online handwritten Hindi numeral recognition system using Support Vector Machines. Co-ordinate points of the input handwritten numeral are collected; various algorithms for pre-processing are applied for normalizing, resampling and interpolating missing points. Angle, curvature along with the x and y coordinates are extracted from the input handwritten numeral. The data obtained is then used for recognition using the kernel functions of SVM. The recognition accuracies are obtained on different schemes of data using the four kernel functions of SVM.

After the invention of computers, a great amount of work has been done in the field of computer human interface. But the problem of exchanging data between human beings and computing machines is still challenging. An enormous research has been done for efficient character recognition and even now research is going on.

i | P a g e ii | P a g e Acknowledgement No volume of words is enough to express my gratitude towards my guide, Mr. Karun. Verma, Assistant. Professor and also the P.G. Coordinator, Computer Science and Engineering Department, Thapar University, who has been very concerned and has aided for all the material essential for the preparation of this thesis report. He has helped me to explore this vast topic in an organized manner and provided me with all the ideas on how to work towards a research-oriented venture. I am also thankful to Dr. Maninder Singh, Head of Department, CSEDr, for the motivation and inspiration that triggered me for the thesis work. I would also like to thank the staff members and my colleagues who were always there in the need of the hour and provided with all the help and facilities, which I required, for the completion of my thesis. Most importantly, I would like to thank my parents and the Almighty for showing me the right direction out of the blue, to help me stay calm in the oddest of the times and keep moving even at times when there was no hope. Sushate Sandal 801032028 iii | P a g e Abstract Mobile ad hoc network (MANET) is an autonomous system of mobile nodes connected by wireless links. Each node operates not only as an end system, but also as a router to forward packets. The nodes are free to move about and organize themselves into a network. These nodes change positions frequently. To accommodate the changing topology special routing algorithms are needed. The main classes of routing protocols are Flat Routing, Hierarchical Routing and Geographic Position assisted Routing.

Bat Algorithm is one of the recently-developed nature-inspired methods in the field of computational intelligence. This work presents the development of the Bat Algorithm based Scheduling Tool (BAST) that used to solve multi-stage multi-machine multi-product scheduling problems. The algorithm takes into account the Just-in-Time production philosophy by aiming to minimise the combination of earliness and tardiness penalty costs. The computational experiment on the BAST was conducted using data obtained from a collaborating company engaged in capital goods industry. The experimental results indicated that the BA performance can be improved up to 8.37% after adopting the appropriate parameters' setting.

Power consumption is a major factor that limits the performance of computers. We survey the "state of the art" in techniques that reduce the total power consumed by a microprocessor system over time. These techniques are applied at various levels ranging from circuits to architectures, architectures to system software, and system software to applications. They also include holistic approaches that will become more important over the next decade. We conclude that power management is a multifaceted discipline that is continually expanding with new techniques being developed at every level. These techniques may eventually allow computers to break through the "power wall" and achieve unprecedented levels of performance, versatility, and reliability. Yet it remains too early to tell which techniques will ultimately solve the power problem.

Cloud computing is emerging as a new paradigm of large-scale distributed computing. It is a framework for enabling convenient, on-demand network access to a shared pool of computing resources. Load balancing is one of the main challenges in ...

Due to significant advances in miniaturization, low power circuit design but reasonably efficient to carry the sensitive information through wireless communication, wireless sensor network (WSN) have attracted attention a lot in recent years.

The rate at which cyber-attacks are increasing globally portrays a terrifying picture upfront. The main dynamics of such attacks could be studied in terms of the actions of attackers and defenders in a cyber-security game. However, currently, little research has taken place to study such interactions. In this paper, we use behavioral game theory and try to investigate the role of certain actions taken by attackers and defenders in a simulated cyber-attack scenario of defacing a website. We choose a Reinforcement Learning (RL) model to represent a simulated attacker and a defender in a 2x4 cyber-security game, where each of the 2 players could take up to 4 actions. A pair of model participants were computationally simulated across 1000 simulations, where each pair played at most 30 rounds in the game. The goal of the attacker was to deface the website and the goal of the defender was to prevent the attacker from doing so. Our results show that the actions taken by both the attackers and defenders are a function of attention paid by these roles to their recently obtained outcomes. It was observed that if attacker pays more attention to recent outcomes, then he is more likely to perform attack actions. We discuss the implication of our results on the evolution of dynamics between attackers and defenders in cyber-security games.

The rate at which cyber-attacks are increasing globally portrays a terrifying picture upfront. The main dynamics of such attacks could be studied in terms of the actions of attackers and defenders in a cyber-security game. However, currently, little research has taken place to study such interactions. In this paper, we use behavioral game theory and try to investigate the role of certain actions taken by attackers and defenders in a simulated cyber-attack scenario of defacing a website. We choose a Reinforcement Learning (RL) model to represent a simulated attacker and a defender in a 2x4 cyber-security game, where each of the 2 players could take up to 4 actions. A pair of model participants were computationally simulated across 1000 simulations, where each pair played at most 30 rounds in the game. The goal of the attacker was to deface the website and the goal of the defender was to prevent the attacker from doing so. Our results show that the actions taken by both the attackers and defenders are a function of attention paid by these roles to their recently obtained outcomes. It was observed that if attacker pays more attention to recent outcomes, then he is more likely to perform attack actions. We discuss the implication of our results on the evolution of dynamics between attackers and defenders in cyber-security games.

Intrusion Detection Systems (IDSs) help in creating cyber situational awareness for defenders by providing recommendations. Prior research in simulation and game-theory has revealed that the presence and accuracy of IDS-like recommendations influence the decisions of defenders and adversaries. In the current paper, we present novel analyses of prior research by analyzing the sequential decisions of defenders and adversaries over repeated trials. Specifically, we developed computational cognitive models based upon Instance-Based Learning Theory (IBLT) to capture the dynamics of the sequential decisions made by defenders and adversaries across numerous conditions that differed in the IDS's availability and accuracy. We found that cognitive mechanisms based upon recency, frequency, and variability helped account for adversarial and defender decisions better than the optimal Nash solutions. We discuss the implications of our results for adversarial-and-defender decisions in the cyber-world.

Cyber-attacks are deliberate attempts by adversaries to illegally access online information of other individuals or organizations. There are likely to be severe monetary consequences for organizations and its workers who face cyber-attacks. However, currently, little is known on how monetary consequences of cyber-attacks may influence the decision-making of defenders and adversaries. In this research, using a cybersecurity game, we evaluate the influence of monetary penalties on decisions made by people performing in the roles of human defenders and adversaries via experimentation and computational modeling. In a laboratory experiment, participants were randomly assigned to the role of "hackers" (adversaries) or "analysts" (defenders) in a laboratory experiment across three between-subject conditions: Equal payoffs (EQP), penalizing defenders for false alarms (PDF) and penalizing defenders for misses (PDM). The PDF and PDM conditions were 10-times costlier for defender participants compared to the EQP condition, which served as a baseline. Results revealed an increase (decrease) and decrease (increase) in attack (defend) actions in the PDF and PDM conditions, respectively. Also, both attack-and-defend decisions deviated from Nash equilibriums. To understand the reasons for our results, we calibrated a model based on Instance-Based Learning Theory (IBLT) theory to the attack-and-defend decisions collected in the experiment. The model's parameters revealed an excessive reliance on recency, frequency, and variability mechanisms by both defenders and adversaries. We discuss the implications of our results to different cyber-attack situations where defenders are penalized for their misses and false-alarms.