Human eye gaze prediction is a challenging problem that has applications in many fields, such as user interface optimization, AR and VR environments, building assistive technologies, optimizing media storage and transmission, and medical applications. Although many studies proposed methods for predicting eye gaze location in images, very few studies have investigated methods of predicting eye gaze location in videos. In this paper, we propose GazeEngine, a deep neural network for predicting eye gaze location in videos. Our model predicts a single gaze location for every frame in the video. In the heart of GazeEngine is a pretrained object detector that extracts salient features from the video frames. Despite the simplicity of the model, preliminary results show that our model performs well for short distance prediction in Coutrot video dataset. Models such as GazeEngine can be applied to predict gaze location to compress video in a continuous streaming setting, where a client is streaming video from server.

This paper presents CLB, a programmable switch-based general-purpose in-network load balancer that can adapt to traffic changes at a very high speed. It uses Weighted-Cost Multipath (WCMP) mechanism for traffic-aware load balancing over many paths at a coarse-grained precision. CLB can be configured to match the load balancing requirements of a wide range of applications at line rate. We have analytically shown that CLB can achieve a bounded response time to traffic changes in the data plane. We implement CLB using the P4 programming language. Our experimental evaluation shows CLB can successfully distribute the incoming load over multiple paths for a given path-weight distribution and leads to performance improvement compared to other load balancing schemes.

IP lookup and packet classification are two core functions of a router. IP lookup involves performing the longest prefix match (LPM) of the destination IP address. Packet classification involves finding the best match in a multi-field ruleset where each field needs an exact or prefix match. ASIC based IP lookup and packet classification are traditionally designed in a register transfer level (RTL) hardware description language (HDL) such as Verilog or VHDL. However, manually writing hardware logic is notoriously complicated and painful. This paper presents a High Level Synthesis (HLS) system named C2RTL. C2RTL generates hardware logic in Verilog RTL directly from IP lookup or packet classification algorithm implemented in C. C2RTL is implemented as a plugin of GCC compiler. It takes an IP lookup or packet classification algorithm (in C) as an input and generates corresponding synthesizable Verilog RTL code for pipelined ASIC. We developed several IP lookup and packet classification algorithms in C2RTL and generated corresponding Verilog RTL. We evaluated the resulting RTL code with OpenROAD EDA.

Routing table lookup is a key function of a router. It involves performing the longest prefix match (LPM) of an IP address. Poptrie, the state-of-the-art trie based routing table lookup, encodes nodes using population counting bitmap. Poptrie uses PopCount CPU instruction which can process only 64 bits at a time. This is why, Poptrie uses 6-bit stride (26= 64). This paper presents an extension of Poptrie named CP-Trie (stands for Cumulative PopCount based Trie) where it stores cumulative PopCount along with population counting bitmap. This enables CP-Trie to process longer stride (e.g. 8-16 bits) at each step. This reduces the number of steps and the number of memory access needed for an IP lookup. The fewer number of steps results in faster lookup. It also results in less power consumption in ASIC. Fewer memory accesses indicate that it requires fewer SRAM blocks in ASIC which results in lower area. This is why, CP-Trie is a more practical solution for pipelined ASIC compared to Poptrie. Our experiments with routing tables from real core routers show that CP-Trie achieves upto 1.43× lookup throughput on a general purpose CPU, but consumes 1.36-1.47× memory compared to Poptrie. We also implemented Poptrie and CP-Trie in a 1 GHz pipelined ASIC. Our physical synthesis report shows that CP-Trie consumes 0.86× power and 0.79× area compared to Poptrie in ASIC.

RCP is a router-assisted congestion control mechanism where routers allocate a feedback rate to each flow. Here sender nodes regulate their sending rates based on the feedback rates received from the routers. An RCP router calculates the feedback rate based on the queue occupancy, spare capacity and the average RTT of the on-going flows. This is why, an RCP router needs to execute a stateful packet processing program in the dataplane. However stateful packet processing in a very highspeed (e.g. Tb/s) router is particularly challenging. Currently there is no real deployment of RCP, to the best of our knowledge. This paper presents an implementation of RCP protocol in a programmable-pipeline based router. We use Domino programming language to implement the dataplane. Domino is a domain specific language that allows us to develop a packet processing program as well as the hardware pipeline on which the packet processing program would be executed. Domino programs are guaranteed to run at line rate. This is the first implementation of RCP using Domino. to the best of our knowledge.

This paper presents a programmable pipeline based Linux router. Here Linux kernel works as a control plane and the programmable pipeline works as a data plane. The Forwarding Information Base (FIB) lookup is performed using SAIL algorithm. SAIL consumes very large memory. Here we use population counting that reduces the memory consumption of SAIL by up to 80%. We have implemented SAIL in Linux kernel. We also have implemented SAIL using Domino programming language. Our implementation shows that a programmable pipeline can execute SAIL at line rate. We have evaluated our implementation with FIBs from real backbone routers. Our experimental results show that SAIL with population counting is suitable for implementing both dataplane and control plane of a high-speed router.

In this paper, we propose Soter a HIPAA compliance distributed personalized authentication mechanism for the Internet of Medical Things (IoMT), based on Multiparty Trust Negotiation. Multiparty Trust Negotiation (MTN)is a promising approach for establishing trust between multiparty (Things)demanding to share resources, by gradually exchanging digital credentials based on personalized Access Control Policies. Previous work has been proposed for authentication in IoMT environment, however the proposed approaches are not the best for IoMT as they are centralized and based on Identity Access Control Policies (IBAC)which have limitations make them unsuitable for open distributed environment such as the IoMT, we explain these limitation on the next section. Soter framework presents a number of innovation features, such as the use of Multiparty Trust Negotiation to establish trust between Medical Things, the support of Virtual Federations (VFs)and the Circle of Trust, and the use of personalized and dynamic access control policies (DACP).

Multiparty Trust Negotiation (MTN) is an emerging methodology for establishing trust between peers by gradually requesting and disclosing digitally signed credentials. However, a MTN strategy must be utilized to search for a successful negotiation based on the Access Control Policies (ACP). Previous work in this area focused on establishing trust between two parties (ATN), yet real life scenarios involved more than two parties in the negotiation process. Even if multiparty is involved, the Access Control Policies (ACP)proposed are either Identity based Access control policies (IBAC)as well as static policies, where the client has no say in decisions, like in the single sign-on systems, Shibboleth [11], and OpenID [14]; or the process is state full and has a high communication cost like in Zhang and Winslett work [9]. This paper provides a formal definition for the multiparty trust negotiation problem and proposes a novel MTN strategy, SafeFlow- x. SafeFlow-x is a stateless, safe, efficient and complete strategy. In the worst case, its communication complexity is O(n), where n is the total number of credentials requested, and its computational complexity is O(k), where k is the total size of peer's ACP when no cyclic dependencies occur.

Interactive video streaming requires very low latency and high throughput. Traditional latency based congestion control algorithm performs poorly in fairness. This results in very poor video quality to adaptive video streaming. Software defined networks (SDN) enables us to solve the problem by designing a network controller in the routers. This paper presents a SDN-centric TCP where sending rate of the network is calculated from the network rather than the host. Routers along the path uses a proportional integral (PI) controller to calculate the sending rate in order to reduce the queuing delay. The routers also divide the available throughput fairly among the flows. The network based controller is found more effective than the sender/receiver based controller in reducing latency and providing fairness. NC-TCP has been designed for interactive video delivery network (VDN) where the interactive video flows compete among themselves. Such differentiated service obviates the need for TCP-friendliness. We have implemented NC-TCP in Linux kernel. We have evaluated NC-TCP in Mininet for an interactive video streaming application. Our experimental results shows that NC-TCP outperforms delay-based congestion control in an interactive VDN.

In this paper, we present a novel Multipoint Video Conferencing (MVC) architecture based on Peer-to-Peer (P2P) platform, such as Gnutella where the Multipoint Control Unit (MCU) is dynamically migrated among the participating peers. The proposed architecture minimizes network traffic and delays by choosing Weighted Center of Graph (WCG) as MCU using an election protocol. We optimize frequency of MCU migration to avoid performance degradation for continuous MCU movement. For large scale video conferencing, single peer MCU can be the bottleneck. So, we distribute the responsibilities of MCU over peers throughout the network applying constraints on bandwidth and computing capacity of each peer. The experimental result shows that the system of Distributed Dynamic MCU (DDMCU) with WCG mostly outperforms systems mediated by Distributed Static MCU (DSMCU).

In this paper we present a study that compares the semantic networks of text comprehension and multimedia comprehension. This comparison is based on the concept learning (CL) model of comprehension. The model, much like artificial neural networks models, mimics the comprehension processes of the human brain. We conducted a human study for the purpose of revealing the semantic variations in comprehending text and comprehending audio-video multimedia. Each participant in the study created a concept semantic network of what they understand, and these networks were processed by the CL-model. The parameters of the CL-model give us insights into the collective learning of the two groups as well as personal performance of each individual. The model metrics are analyzed to reveal quantitative and qualitative differences. The combination of computational modeling of comprehension with semantic networks analysis, makes us able to measure comprehension performance of reader and watchers in a way that was not possible before. Some of the important results that we found indicate that textual media provided easier integration of newly learned concepts with background information. At the same time, we found that recognizing an overwhelming number of concepts is easier with audio-video multimedia. The presented results are important for media creators and educators, as well as artificial intelligence scientists who aim at creating systems that resemble human learning. Similar to the way biology inspired statistical learning algorithms, studying cognitive tasks, such as comprehension, can help us understand human behavior and build systems that imitate human learning.

In this paper, we investigate a novel Multipoint Video Conferencing (MVC) architecture potentially suitable for Peer-to-Peer (P2P) platform, such as Gnutella. In particular, we present an election protocol (extension to Gnutella) where the Multipoint Control Unit (MCU) of the MVC is dynamically migrated among peers when new peer joins or leaves. Simulation result shows that this improves overall conferencing performance compared to the system with static MCU by minimizing total traffic, individual node hotness, and video composition delay.

In this paper we explore automated verification of electronic medical record (EMR) transaction for compliance with a regulatory regimen such as HIPAA. We present an approach based on modeling the conceptual space of HIPAA. The clauses of HIPAA legal text is then converted into a disambiguated decision tree (DDT) rules precisely identifying the allowed, dis-allowed and prescribed actions per work flow request type. Given any EMR query the DT then enables one not only to verify the compliance as well as provide complete release guidance as prescribed by HIPAA, generate explanation and audit.

One of the ways of improving quality of Software Engineering (SE) education is to develop mechanisms for quantitatively analyzing SE courses taught in universities. We propose a new multidimensional approach for analyzing SE courses. Two aspects are important for this analysis. One is what is covered (how much of the total SE knowledge space is covered in the course)? And second is how is it covered (how effectively SE topics and subtopics are covered in lectures)? We call the first one structural analysis and second one pedagogical analysis. Techniques and metrics for structural and pedagogical analysis of SE courses are presented in this paper together with results from implementation of the work on four SE courses.

The goal of this paper is to predict future horizontal eye movement trajectories within a specified time interval. To achieve this goal a linear horizontal oculomotor plant mechanical model is developed. The model consists of the eye globe and two extraocular muscles: lateral and medial recti. The model accounts for such anatomical properties of the eye as muscle location, elasticity, viscosity, eye-globe rotational inertia, muscle active state tension, length tension and force velocity relationships. The mathematical equations describing the oculomotor plant mechanical model are transformed into a Kalman filter form. Such transformation provides continuous eye movement prediction with a high degree of accuracy. The model was tested with 21 subjects and three multimedia files. Practical application of this model lies with direct eye gaze input and interactive displays systems as a method to compensate for detection, transmission and processing delays.

This article designs a real-time perceptual compression system (RTPCS) based on eye-gaze-position analysis. Our results indicate that the eye-gaze-position containment metric provides more efficient and effective evaluation of an RTPCS than the eye fixation containment. The presented RTPCS is designed for a network communication scenario with a feedback loop delay. The proposed RTPCS uses human visual system properties to compensate for the delay and to provide high ratios of multimedia compression.

Congestion control for compressed audio is a nontrivial pursuit. Audio perception is highly susceptible to disturbance in temporal quality. We recently proposed and implemented a novel symbiotic perceptual audio streaming mechanism, which receives fast feedback from network and responds to impending congestion with advanced auditory perception based adaptation techniques and supports near flawless sound. The design is based on recently proposed iTCP. It enables a new principle where encoder trade-off fidelity for accurate audio temporal quality using fast network event feedback. We have tested it live over the Internet. In this paper, we share the performance of this system and report observed improvements in time-bounded streaming audio.

Interactive Transparent Networking has been proposed to support a new generation of symbiotic applications that require advance interaction with the Network. In this mode advanced applications can subscribe to state- space event based feed from network protocol local end- points by interactive version of the network protocols. This enables a whole new range of high performance extensible adaptation which requires low-time constant feedback. In this paper, we show how classical TCP can be extended to support long delay high capacity pipe. We demonstrate a system that mimics FAST TCP however, without the usual extensive reengineering required by the original. It seems matching performance can be achieved with protocol interactivity.

In this paper we present a method to represent knowl-edge associated with a course and observe clustering of concepts for educational resources associated with that course. Course knowledge can be represented in the form of hierarchical prerequisite relation based weighted ontology. A novel approach for selectively processing relevant parts of the ontology is given. Design and analytical information extraction from educational re-sources is possible using this approach. The applica-bility of this method is not limited to the domain of education and can be used to analyze artifacts from any domain in which knowledge can be represented as a structured hierarchy.

Online transactional activities that involve establishment of trust between participating individual seem to require a reputation function for the reputation estimation framework (REF). They are often vulnerable to various kinds of attacks. Also it seems we do not evaluate reputation in the same way in all situations. Using Occam's razor we propose a generalized set-theoretic reputation function with customizable components that can be changed to meet the reputation requirements in wide variety of reputation assessment scenarios. Further we identify several canonical classes of the functions. The resilience of the framework is then analyzed by subjecting it to various reputation attacks such as gang attacks, vendetta and Dr Jekyll & Mr. Hyde.

Human eyes have limited perception capabilities. Only 2 degrees of our 180 degree vision field provide the highest quality of perception. Due to this fact the idea of perceptual attention focus emerged to allow a visual content to be changed in a way that only part of the visual field where a human attention is directed to is encoded with a high quality. The image quality in the periphery can be reduced without a viewer noticing it. This compression approach allows a significant decrease in bit-rate for a video stream, and in the case of the 3D stream rendering, it decreases the computational burden. A number of previous researchers have investigated the topic of real-time perceptual attention focus but only for a single viewer. In this paper we investigate a dynamically changing multi-viewer scenario. In this type of scenario a number of people are watching the same visual content at the same time. Each person is using eye-tracking equipment. The visual content (video, 3D stream) is sent through a network with a large transmission delay. The area of the perceptual attention focus is predicted for the viewers to compensate for the delay value and identify the area of the image which requires highest quality coding.

This paper describes an approach that can facilitate composition of course materials by machine assistance over the Web using semantic Web technology. A semantic topic graph (STG) that encodes the conceptual knowledge metadata space is used as the front-end to the course resources. STG provides ontologies and properties of the concepts space and course materials are then connected with respect to their relation to this concept space. A course composer system is demonstrated which can intelligently compose various course materials based on pedagogical consideration. Some interesting comparisons of human guided machine-composed courses with real human composed courses in use in US graduate programs are shown

Online communities are the latest phenomena on the Internet. At the heart of each community lies a social network. In this paper, we show a generalized framework to understand and reason in social networks. Previously, researchers have attempted to use inference-specific type of relationships. In this paper, we propose a framework to represent and reason with general case of social relationship network in a formal way. We call it relationship algebra. This paper presents this algebra and then shows how this algebra can be used for various interesting computations on a social network weaved in the virtual communities

Human eyes have limited perception capabilities; for example, only 2 degrees of our 140 degree vision field provide the highest quality of perception. Due to this fact the idea of perceptual focus emerged to allow a visual content to be changed in a way that only part of the visual field where a human gaze is directed is encoded with a high quality. The image quality in the periphery can be reduced without a viewer noticing it. This compression approach allows a significant decrease in the number of bits required for image encoding, and in the case of the 3D image rendering, it decreases the computational burden. A number of previous researchers have investigated the topic of perceptual focus but only for a single viewer. In our research we investigate a dynamically changing multi-viewer scenario. In this type of scenario a number of people are watching the same visual content at the same time. Each person has his/her own perceptual focus area which changes over time. The visual content is sent through a network with a fixed delay/lag which provides an additional challenge to the whole scheme. The goal of our work was to investigate and develop a method of multi-viewer perceptual focus zones adaptation for real-time media perceptual compression and transmission. In our research we also look into the impact that such a method can have on transmission bandwidth and computational burden reduction.

Testing and evaluation is an integral part of the learning process. Educators have often tried to device methods for design of test-ware. Intelligent design and compilation of test-ware is a very interesting problem with immense applications. This research aims at automating the process of intelligent design of test-ware by providing qualitative assessment of questions. In this attempt, we provide some synthetic parameters for the evaluation of question in its concept space. The parameters are tested in some real world scenarios and intuitive inferences are deduced predicting the performance of the parameters. It is observed that the difficulty of a question is often a function of the concepts it tests. Concept knowledge can be represented in the form of linked concepts in semantic nets, the links representing the relationships between the concepts. If this directed graph is known, complexity of a question can be computed by synthetic means.

Loss-free handoff in mobile networks is an extensive research area. Mobile IP (MIP) provided a solution to enable a mobile node to roam from one location to another while maintaining its network level connectivity. However, handoff latencies and longer triangular routing paths in MIP can severely degrade communication performance and in particular cripple connection oriented protocols like TCP. In this paper we propose an alternate approach for robust mobility. The scheme is based on the principle of 'interactive transparent networking' where all networking layers remain lightweight but are engineered for interactivity. This would allow principle intelligent actions to be performed at the application layer. With protocol interactivity we demonstrate a novel scheme that switches IP address in the TCP/IP stack on both end-points and perform loss-free rapid handoff. The scheme offers not only loss-free handoff, but also offers several fundamental system advantages; (i) it does not impose any changes on original network protocols or their dynamics, and (ii) it fully adheres to the end-to-end principle and do not require intermediary nodes as in MIP. We have achieved a real implementation of the scheme on FreeBSD and tested the real system over Internet with voice traffic. We show that this scheme can dramatically reduce handoff latency and improve TCP performance by offering shorter routes with loss-free handoffs and smooth, low-jitter voice stream.

MOTIVATION: The accumulation of genomic alterations is an important process in tumor formation and progression. Comparative genomic hybridization performed on cDNA arrays (cDNA aCGH) is a common method to investigate the genomic alterations on a genome-wide scale. However, when detecting low-level DNA copy number changes this technology requires the use of noise reduction strategies due to a low signal to noise ratio. RESULTS: Currently a running average smoothing filter is the most frequently used noise reduction strategy. We analyzed this strategy theoretically and experimentally and found that it is not sensitive to very low level genomic alterations. The presence of systematic errors in the data is one of the main reasons for this failure. We developed a novel algorithm which efficiently reduces systematic noise and allows for the detection of low-level genomic alterations. The algorithm is based on comparison of the biological relevant data to data from so-called self-self hybridizations, additional experiments which contain no biological information but contain systematic errors. We find that with our algorithm the effective resolution for +/-1 DNA copy number changes is about 2 Mb. For copy number changes larger than three the effective resolution is on the level of single genes.

No abstract available.

This work presents a fast near linear run-time system for customizable criteria based packet forwarding and route lookup. It presents the case for deadline line based packet forwarding. This system has been designed after the constraint resource planning (CRP) methodology for algorithm designs. The resulting near-optimum algorithm operates within an extensible CRP shell and uses two pluggable scheduling and forwarding heuristics to produce near-optimal performance. We present the shell architecture, the algorithm, and the simulation results of this system for dead-line based packet forwarding.

Approximately 2 degrees in our 140 degree vision span has sharp vision. Many researchers have been fascinated by the idea of eye-tracking integrated perceptual compression of an image or video, yet any practical system has yet to emerge. The unique challenge presented by real time perceptual video streaming is how to handle the fast nature of the human eye and provide its integration with computationally intensive video transcoding scheme. The delay introduced by video transmission in the network presents a difficulty. This delay creates a problem when we try to use information about eye movements for perceptual encoding. In this paper we discuss a new approach to the eye-tracker based video compression. Rather than relying on the point of gaze, this novel scheme tracks a vicinity of interest and offers a prediction mechanism for eye movements. The described system compensates the interim eye movements between the sampling and actual coding. The proposed scheme can be applied to a large variety of today's video compression standards. We have developed an eye gaze-aware MPEG-2 transcoder that can perceptually re-encode a live video stream in real time. The experiments we have conducted illustrate the substantial impact this integrated prediction method has on perceptual video compression and bit-rate reduction.

The possibility of perceptual compression using live eye-tracking has been anticipated for some time by many researchers. Among the challenges of real-time eye-gaze based perceptual video compression is how to handle the fast nature of eye movements with a relative complexity of video transcoding and also take into the account a delay associated with transmission in the network. Such delay requires an additional consideration in perceptual encoding because it increases the size of the area that requires high quality coding. In this paper we present a hybrid scheme, one of the first to our knowledge, which combines eye-tracking with fast in-line scene analysis to drastically narrow down the high acuity area without the loss of eye-gaze containment.

The paper presents an interesting study that how the user surfing behavior with respect to the organization of a Web space affects the performance of a prefetch enabled proxy. We have conducted experiments based on a novel hyperspace aware prefetch proxy and have studied the prefetch performance on tree and complete graph. The study assesses the system's responsiveness as well as the background loads for various user interaction duration, surfing and prefetch sequences. The results show that awareness about the Web space and surfing sequence can greatly help in improving prefetch performance. Schemes such as these also conversely may enable professional content developers to create prefetch friendly collection for increased site responsiveness.

Interactivity in transport protocol can greatly benefit transport friendly applications generating streaming traffic. Recently we have developed the TCP interactive, which can provide event notification to the subscriber of its communication service. This is operationally state equivalent to the conventional TCP except applications can optionally subscribe, receive, and in real-time react to selected local end-point events. This simple extension opens the horizon for a spectrum of smart application level solutions to be realized for many of the current hard problems including congestion control for time sensitive elastic traffic. We have recently implemented and tested the real system on the active network (ABone) testbed for video streaming to worldwide sites. In this paper we share the performance of this system and report potential dramatic improvement in time-bounded streaming traffic.

The exploration of high dimensional gene expression microarray data demands powerful analytical tools. Our data mining software, visual data analyzer (VISDA) for cluster discovery, reveals many distinguishing patterns among gene expression profiles. The model-supported hierarchical data exploration tool has two complementary schemes: discriminatory dimensionality reduction for structure-focused data visualization, and cluster decomposition by probabilistic clustering. Reducing dimensionality generates the visualization of the complete data set at the top level. This data set is then partitioned into subclusters that can consequently be visualized at lower levels and if necessary partitioned again. These approaches produce different visualizations that are compared against known phenotypes from the microarray experiments. For class prediction on cancers using miroarray data, multilayer perceptrons (MLPs) are trained and optimized, whose architecture and parameters are regularized and initialized by weighted Fisher criterion (wFC)-based discriminatory component analysis (DCA). The prediction performance is compared and evaluated via multifold cross-validation.

Interactivity in transport protocol can greatly benefit transport friendly applications. We envision if a transport mechanism, which is interactive and can provide event notification about network state to the subscriber of its communication service, than a wide range of solution to many of today's hard network problem can be instituted. Recently we have implemented this concept system as a new TCP kernel on FreeBSD called TCP interactive and a novel symbiotic MPEG-2 full logic transcoder. In this paper we share the results of the TCP interactive performance experiment and show potential dramatic improvement in time-bounded video delivery.

Interception of a data stream is central to any intelligent and dynamic processing of Web information. It is perhaps as fundamental to Internet services' overall architecture as the design of disk scheduling to the conventional machine architecture. We discuss an IPv6 based indexing protocol that can facilitate random access into multilevel hierarchically encoded content streams and provide serious performance boost to Web content processing.

We explore how the structure of Webspace can be used for accelerated Web prefetch. We have conducted experiments based on a novel hyperspace aware prefetch proxy and have studied the prefetch performance on several dominant hyperspace patterns. We assess the system's responsiveness and background loads for various user interaction duration, surfing and prefetch sequences. The results show that Webspace awareness can help in improving prefetch performance. We also provide an interesting insight toward a framework where the professional content developers can gain more control towards authoring prefetch friendly collection for increased site responsiveness.

Composite hypermedia documents are fast becoming ubiquitous in the Web. More Websites are effectively serving hypermedia however with hypertext technology. We explore a novel segment-based background prefetch technique that utilizes the rendering properties of various individual media elements and their interdependencies. The technique can accelerate surfing in composite hypermedia.

Object based bit allocation can result in significant improvement in the perceptual quality of relatively low bit-rate video. In this paper we describe a novel content aware video transcoding technique that can accept high-level description of video objects and extract them from incoming video stream and use it for perceptual encoding based extreme video downscaling.

In this paper we discuss architecture for network aware adaptive systems for next generation networks. We present in the context of a novel cognizant video transcoding system, which is capable of negotiating local network state based rate and let the video propagate over extreme network with highly asymmetric link and node capacities utilizing knowlege about the network, content protocol and the content itself.

The authors present a potential new approach of application integrated congestion management. It is applicable for time-sensitive traffic and is based on the principle of direct protocol interactivity. In contrast to classical transport protocols, we envision a transport mechanism which is interactive and can provide event notification to the subscriber of its communication service. We then show a network aware adaptive MPEG-2 video transcoding mechanics, which directly interacts with the protocol and adjusts its production in sync with the impairment events in the transport layer. The authors present the novel symbiotic mechanics, and share the performance result observed on the MPEG-2 video stream carried by this symbiosis implementation. We report dramatic improvement in time-conformant video delivery by such protocol interactivity.

The paper presents a scheme for reducing memory space of a holographic associative memory for content-based learning, searching and retrieval of sparse patterns. Multidimensional holographic associative memory developed on the properties of complex valued Riemann space is one of the most promising models of associative memory. It has demonstrated the unique ability to perform dynamically localizable sub-pattern matching, without requiring to learn each individual sub-patterns. The correlation space of the sparse patterns, is also sparse in information, but representationally dense. Therefore, holograph of sparse patterns (such as images) becomes extremely large. In this paper we describe a holographic memory model which can prune a holograph by several fold. The resulting holographic model also simultaneously increases the encoding, searching and decoding speed.

This paper presents the performance analysis (capacity and retrieval accuracy) of multidimensional holographic associative memory (MHAC). MHAC has the unique ability to retrieve pattern-associations with changeable attention. In attention actuated retrieval the user can dynamically select any subset of the elements in the example query pattern and expect the memory to confine its associative match only within the specified field of attention. Existing artificial associative memories lack this ability. Also most of these models need at least 50% of bits in the input pattern to be correct for successful retrieval. MHAC, with the unique ability of localizable attention, can retrieve information correctly even with cues as small as 10% of the query frame. This paper investigates the performance of MHAC in attention actuated retrieval both analytically and experimentally. Besides confirmation, the experiments also identify an operational range space (ORS) for this memory within which various attention based applications can be built with a performance guarantee.

This paper presents a dynamically adaptive stabilization scheme for parallel matrix computation. The scheme performs automatic error detection and correction through inserting redundant, but concurrent tracer computations within the folds of the regular computation. It also eliminates the costly row interchange used in classical pivoting. A fault-tolerant double wavefront matrix algorithmfor a MIMD array multi-processor with toroidal inter connection has been designed to demonstrate the strength of the proposed scheme. This algorithm can compute: i) matrix inverse ii) solution vector to the linear system and Hi) predetermined linear combination of the solution vector from identical algorithmic framework. This efficient tri-solution algorithm excels most other known methods in parallel performance.

Design extraction of an unfamiliar system is a complex cognitive task. This paper presents an approach that can help human expert in exploring a large and complex code information space of unfamiliar software. It provides her/him a platform to access the code information with flexible, fine and delicate control over volume and composition of the accessed information sub-space. The proposed approach integrates two forms of abstraction. First, it helps one to comprehend complexity of the code information space by allowing the explorer to investigate the system from numerous (combinatorial) coherent perspectives. In the second level, it helps one to overcome scale of the information space by allowing the explorer to compress or expand any composition of its sub-spaces. This new approach, called the adiabatic multi-perspective (AMP) approach to program abstraction, is founded on a symmetrical dual hierarchical (SDH) organization of code information space and a novel formalism for abstract dependency analysis (ADA), which is also one of the first formalisms to perform complete program dependency analysis on abstract program models.>

Technically, because of the limited dimensions of the presentation media, and psychologically, because of the limited human attention span, graphical program visualization in computer assisted reverse engineering becomes increasingly difficult with scale. This problem becomes compounded when multi-perspective visualization is taken into consideration. In this research we present a system called HVIEW, which attempts to overcome the problem of scale by logical decomposition. HVIEW can present a graphical visual of a computer program at various abstraction levels from diverse perspectives as interactively requested by the user. HVIEW logically formulates abstract function and data objects to construct compressed immediate program representations and organizes them into independent data and function hierarchies. Whereas, these hierarchies provide the mechanism of flexible abstraction, a user specifiable query filter creates an assortment of perspectives on the rendered visual of the modelled code.>