Showing papers by "INESC-ID published in 2007"

YEASTRACT-DISCOVERER: new tools to improve the analysis of transcriptional regulatory associations in Saccharomyces cerevisiae

Abstract: The Yeast search for transcriptional regulators and consensus tracking (YEASTRACT) information system (www.yeastract.com) was developed to support the analysis of transcription regulatory associations in Saccharomyces cerevisiae. Last updated in September 2007, this database contains over 30 990 regulatory associations between Transcription Factors (TFs) and target genes and includes 284 specific DNA binding sites for 108 characterized TFs. Computational tools are also provided to facilitate the exploitation of the gathered data when solving a number of biological questions, in particular the ones that involve the analysis of global gene expression results. In this new release, YEASTRACT includes DISCOVERER, a set of computational tools that can be used to identify complex motifs over-represented in the promoter regions of co-regulated genes. The motifs identified are then clustered in families, represented by a position weight matrix and are automatically compared with the known transcription factor binding sites described in YEASTRACT. Additionally, in this new release, it is possible to generate graphic depictions of transcriptional regulatory networks for documented or potential regulatory associations between TFs and target genes. The visual display of these networks of interactions is instrumental in functional studies. Tutorials are available on the system to exemplify the use of all the available tools.

FearNot!: an emergent narrative approach to virtual dramas for anti-bullying education

Abstract: FearNot! is a story-telling application originally created in the EU FP5 project VICTEC and now extended in the FP6 project eCIRCUS [eCIRCUS 07]. It has applied ideas from Forum Theatre [Boal 79] to the domain of education against bullying. In Forum Theatre, sections of an audience take responsibility for a specific character in the unfolding drama, played by an actor who always stays in role. Episodes in which the actors improvise within an overall narrative framework are broken by interaction sections in which the audience sections talk over with 'their' character what they should do in the next dramatic segment. The actor is free to reject advice that seems incompatible with their role, and may also suspend a dramatic episode if it seems necessary to get further advice.

Frustratingly Hard Domain Adaptation for Dependency Parsing

Abstract: We describe some challenges of adaptation in the 2007 CoNLL Shared Task on Domain Adaptation. Our error analysis for this task suggests that a primary source of error is differences in annotation guidelines between treebanks. Our suspicions are supported by the observation that no team was able to improve target domain performance substantially over a state of the art baseline.

A combined component approach for finding collection-adapted ranking functions based on genetic programming

Abstract: In this paper, we propose a new method to discover collection-adapted ranking functions based on Genetic Programming (GP). Our Combined Component Approach (CCA)is based on the combination of several term-weighting components (i.e.,term frequency, collection frequency, normalization) extracted from well-known ranking functions. In contrast to related work, the GP terminals in our CCA are not based on simple statistical information of a document collection, but on meaningful, effective, and proven components. Experimental results show that our approach was able to outper form standard TF-IDF, BM25 and another GP-based approach in two different collections. CCA obtained improvements in mean average precision up to 40.87% for the TREC-8 collection, and 24.85% for the WBR99 collection (a large Brazilian Web collection), over the baseline functions. The CCA evolution process also was able to reduce the overtraining, commonly found in machine learning methods, especially genetic programming, and to converge faster than the other GP-based approach used for comparison.

Automated smoother for the numerical decoupling of dynamics models.

Abstract: Structure identification of dynamic models for complex biological systems is the cornerstone of their reverse engineering. Biochemical Systems Theory (BST) offers a particularly convenient solution because its parameters are kinetic-order coefficients which directly identify the topology of the underlying network of processes. We have previously proposed a numerical decoupling procedure that allows the identification of multivariate dynamic models of complex biological processes. While described here within the context of BST, this procedure has a general applicability to signal extraction. Our original implementation relied on artificial neural networks (ANN), which caused slight, undesirable bias during the smoothing of the time courses. As an alternative, we propose here an adaptation of the Whittaker's smoother and demonstrate its role within a robust, fully automated structure identification procedure. In this report we propose a robust, fully automated solution for signal extraction from time series, which is the prerequisite for the efficient reverse engineering of biological systems models. The Whittaker's smoother is reformulated within the context of information theory and extended by the development of adaptive signal segmentation to account for heterogeneous noise structures. The resulting procedure can be used on arbitrary time series with a nonstationary noise process; it is illustrated here with metabolic profiles obtained from in-vivo NMR experiments. The smoothed solution that is free of parametric bias permits differentiation, which is crucial for the numerical decoupling of systems of differential equations. The method is applicable in signal extraction from time series with nonstationary noise structure and can be applied in the numerical decoupling of system of differential equations into algebraic equations, and thus constitutes a rather general tool for the reverse engineering of mechanistic model descriptions from multivariate experimental time series.

The Titech large vocabulary WFST speech recognition system

Abstract: In this paper we present evaluations on the large vocabulary speech decoder we are currently developing at Tokyo Institute of Technology. Our goal is to build a fast, scalable, flexible decoder to operate on weighted finite state transducer (WFST) search spaces. Even though the development of the decoder is still in its infancy we have already implemented a impressive feature set and are achieving good accuracy and speed on a large vocabulary spontaneous speech task. We have developed a technique to allow parts of the decoder to be run on the graphics processor, this can lead to a very significant speed up.

A Wireless Sensor MAC Protocol for Bursty Data Traffic

Abstract: This paper proposes MH-MAC, a new MAC protocol for wireless sensor networks capable of handling applications that generate infrequent huge peaks of traffic. Existing protocols are not adapted to this kind of applications. Asynchronous protocols are energy efficient for the long inactive periods, but fail to cope with the bandwidth and latency requirements of the traffic peaks when more than two nodes are sending data to a common sink. Synchronous protocols that support contention free slots provide good throughput for handling the load peaks, but consume unnecessary energy maintaining clocks synchronized for very long idle periods. MH-MAC is a multimode hybrid protocol that can be configured by the application to run in asynchronous mode or in synchronous mode, with or without contention, providing the best possible trade-off. MH-MAC is a single-hop MAC, which supports multi-hop applications through a cross-layering API. The paper includes simulation results with the energy consumption, latency and throughput for the operation modes of MH-MAC, showing the asynchronous-synchronous trade-offs and the state transition overhead.

Local Renyi entropic profiles of DNA sequences

Abstract: In a recent report the authors presented a new measure of continuous entropy for DNA sequences, which allows the estimation of their randomness level. The definition therein explored was based on the Renyi entropy of probability density estimation (pdf) using the Parzen's window method and applied to Chaos Game Representation/Universal Sequence Maps (CGR/USM). Subsequent work proposed a fractal pdf kernel as a more exact solution for the iterated map representation. This report extends the concepts of continuous entropy by defining DNA sequence entropic profiles using the new pdf estimations to refine the density estimation of motifs. The new methodology enables two results. On the one hand it shows that the entropic profiles are directly related with the statistical significance of motifs, allowing the study of under and over-representation of segments. On the other hand, by spanning the parameters of the kernel function it is possible to extract important information about the scale of each conserved DNA region. The computational applications, developed in Matlab m-code, the corresponding binary executables and additional material and examples are made publicly available at http://kdbio.inesc-id.pt/~svinga/ep/ . The ability to detect local conservation from a scale-independent representation of symbolic sequences is particularly relevant for biological applications where conserved motifs occur in multiple, overlapping scales, with significant future applications in the recognition of foreign genomic material and inference of motif structures.

Efficient hybrid DCT-domain algorithm for video spatial downscaling

Abstract: A highly efficient video downscaling algorithm for any arbitrary integer scaling factor performed in a hybrid pixel transform domain is proposed. This algorithm receives the encoded DCT coefficient blocks of the input video sequence and efficiently computes the DCT coefficients of the scaled video stream. The involved steps are properly tailored so that all operations are performed using the encoding standard block structure, independently of the adopted scaling factor. As a result, the proposed algorithm offers a significant optimization of the computational cost without compromising the output video quality, by taking into account the scaling mechanism and by restricting the involved operations in order to avoid useless computations. In order to meet any system needs, an optional and possible combination of the presented algorithm with high-order AC frequency DCT coefficients discarding techniques is also proposed, providing a flexible and often required complexity scalability feature and giving rise to an adaptable tradeoff between the involved scalable computational cost and the resulting video quality and bit rate. Experimental results have shown that the proposed algorithm provides significant advantages over the usual DCT decimation approaches, both in terms of the involved computational cost, the output video quality, and the resulting bit rate. Such advantages are even more significant for scaling factors other than integer powers of 2 and may lead to quite high PSNR gains.

Modelling heavy tails and asymmetry using ARCH -type models with stable Paretian distributions

Abstract: Several approaches have been considered to model the heavy tails and asymmetric effect on stocks returns volatility. The most commonly used models are the Exponential Generalized AutoRegressive Conditional Heteroskedasticity (EGARCH), the Threshold GARCH (TGARCH), and the Asymmetric Power ARCH (APARCH) which, in their original form, assume a Gaussian distribution for the innovations. In this paper we propose the estimation of all these asymmetric models on empirical distributions of the Standard & Poor’s (S&P) 500 and the Financial Times Stock Exchange (FTSE) 100 daily returns, assuming the Student’s t and the stable Paretian with (α < 2) distributions for innovations. To the authors’ best knowledge, analysis of the EGARCH and TGARCH assuming innovations with α-stable distribution have not yet been reported in the literature. The results suggest that this kind of distributions clearly outperforms the Gaussian case. However, when α-stable and Student’s t distributions are compared, a general conclusion should be avoided as the goodness-of-fit measures favor the α-stable distribution in the case of S&P 500 returns and the Student’s t distribution in the case of FTSE 100.

XIS-UML Profile for eXtreme Modeling Interactive Systems

Abstract: The first version of the XIS profile addressed the development of interactive systems by defining models oriented only towards how the system should perform tasks. However, issues such as user-interface layouts, or the capture of interaction patterns, were not addressed by the profile, but only by the source-code generation process. This originated systems that, although functional, were considered by end-users as "difficult to use". In this paper we present the second version of the XIS UML profile, which is now a crucial component of the ProjectIT research project. This profile follows the "separation of concerns" principle by proposing an integrated set of views that address the various issues detected with the previous version of XIS. In addition, this profile also promotes the usage of extreme modeling, by relying on the extensive use of model-to-model transformation templates that are defined to accelerate the model development tasks

Semi-supervised single-label text categorization using centroid-based classifiers

Abstract: In this paper we study the effect of using unlabeled data in conjunction with a small portion of labeled data on the accuracy of a centroid-based classifier used to perform single-label text categorization. We chose to use centroid-based methods because they are very fast when compared with other classification methods, but still present an accuracy close to that of the state-of-the-art methods. Efficiency is particularly important for very large domains, like regular news feeds, or the web.We propose the combination of Expectation-Maximization with a centroid-based method to incorporate information about the unlabeled data during the training phase. We also propose an alternative to EM, based on the incremental update of a centroid-based method with the unlabeled documents during the training phase.We show that these approaches can greatly improve accuracy relatively to a simple centroid-based method, in particular when there are very small amounts of labeled data available (as few as one single document per class).Using one synthetic and three real-world datasets, we show that, if the initial model of the data is sufficiently precise, using unlabeled data improves performance. On the other hand, using unlabeled data degrades performance if the initial model is not precise enough.

IMPROVE : An innovative application for collaborative mobile mixed reality design review

Abstract: In this paper we introduce an innovative application aiming at combining large, tablet-based and headmounted displays for collaborative mobile mixed reality design reviews. Our research and development is motivated by two use scenarios: automotive and architectural design review involving real users from Page/Park architects and Elasis FIAT. Our activities are supported by the EU IST project IMPROVE. It covers activities in the areas of: HMD development using unique OLED technology, markerless tracking, augmented reality rendering, image calibration for large tiled displays, collaborative tablet-based and projection wall oriented interaction and stereoscopic video streaming for mobile users. The paper gives an overview of the hardware and software developments within IMPROVE and concludes with results from first user test.

Optimization of area in digital FIR filters using gate-level metrics

Abstract: In the paper, we propose a new metric for the minimization of area in the generic problem of multiple constant multiplications, and demonstrate its effectiveness for digital FIR filters. Previous methods use the number of required additions or subtractions as a cost function. We make the observation that not all of these operations have the same design cost. In the proposed algorithm, a minimum area solution is obtained by considering area estimates for each operation. To this end, we introduce accurate hardware models for addition and subtraction operations in terms of gate-level metrics, under both signed and unsigned representations. Our algorithm not only computes the best design solution among those that have the same number of operations, but is also able to find better area solutions using a non-minimum number of operations. The results obtained by the proposed exact algorithm are compared with the results of the exact algorithm designed for the minimum number of operations on FIR filters and it is shown that the area of the design can be reduced by up to 18%.

Caravela: A Novel Stream-Based Distributed Computing Environment

Abstract: Distributed computing implies sharing computation, data, and network resources around the world. The Caravela environment applies a proposed flow model for stream computing on graphics processing units that encapsulates a program to be executed in local or remote computers and directly collects the data through the memory or the network.

LABRADOR: Efficiently publishing relational databases on the web by using keyword-based query interfaces

Abstract: A vast amount of valuable information, produced and consumed by people and institutions, is currently stored in relational databases. For many purposes, there is an ever increasing demand for having these databases published on the Web, so that users can query the data available in them. An important requirement for this to happen is that query interfaces must be as simple and intuitive as possible. In this paper we present LABRADOR, a system for efficiently publishing relational databases on the Web by using a simple text box query interface. The system operates by taking an unstructured keyword-based query posed by a user and automatically deriving an equivalent SQL query that fits the user's information needs, as expressed by the original query. The SQL query is then sent to a DBMS and its results are processed by LABRADOR to create a relevance-based ranking of the answers. Experiments we present show that LABRADOR can automatically find the most suitable SQL query in more than 75% of the cases, and that the overhead introduced by the system in the overall query processing time is almost insignificant. Furthermore, the system operates in a non-intrusive way, since it requires no modifications to the target database schema.

I Know What I Did Last Summer: Autobiographic Memory in Synthetic Characters

Abstract: According to traditional animators, the art of building believable characters resides in the ability to successfully portray a character's behaviour as the result of its internal emotions, intentions and thoughts. Following this direction, we want our agents to be able to explicitly talk about their internal thoughts and report their personal past experiences. In order to achieve it, we look at a specific type of episodic long term memory. This paper describes the integration of Autobiographic Memory into FAtiMA, an emotional agent architecture that generates emotions from a subjective appraisal of events.

Dynamic language modeling for a daily broadcast news transcription system

Abstract: When transcribing Broadcast News data in highly inflected languages, the vocabulary growth leads to high out-of-vocabulary rates. To address this problem, we propose a daily and unsupervised adaptation approach which dynamically adapts the active vocabulary and LM to the topic of the current news segment during a multi-pass speech recognition process. Based on texts daily available on the Web, a story-based vocabulary is selected using a morpho-syntatic technique. Using an Information Retrieval engine, relevant documents are extracted from a large corpus to generate a story-based LM. Experiments were carried out for a European Portuguese BN transcription system. Preliminary results yield a relative reduction of 65.2% in OOV and 6.6% in WER.

Clear-PEM: A PET imaging system dedicated to breast cancer diagnostics

Abstract: The Clear-PEM scanner for positron emission mammography under development is described. The detector is based on pixelized LYSO crystals optically coupled to avalanche photodiodes and readout by a fast low-noise electronic system. A dedicated digital trigger (TGR) and data acquisition (DAQ) system is used for on-line selection of coincidence events with high efficiency, large bandwidth and small dead-time. A specialized gantry allows to perform exams of the breast and of the axilla. In this paper we present results of the measurement of detector modules that integrate the system under construction as well as the imaging performance estimated from Monte Carlo simulated data.

An FPGA Implementation for a Kalman Filter with Application to Mobile Robotics

Abstract: The problem of simultaneous localization and mapping has been studied by the mobile robotics scientific community over the last two decades. Most solutions for this problem are based on probabilistic theory in order to represent the uncertainty in robot perception and action. One of the most efficient probabilistic methods is the extended Kalman filter (EKF). However, the EKF demands a considerable amount of computing power and is usually processed by high-end laptops coupled to the robots. In this work, we present an implementation of the EKF targeting an embedded system based on an FPGA device. In order to improve performance, our approach combines a softcore processor with customized hardware. We present experiments with four different FPGA implementations, being the first purely based on software, the second using custom instruction logic directly connected to the processor's ALU, the third using hardware accelerators connected to the processor's data bus, and finally the fourth combining those two hardware/software solutions. For the experiments conducted, the results obtained with a small addition of hardware resources permitted to increase from 2times to 4times the performance of the global system.

New Normalized LMS Algorithms Based on the Kalman Filter

Abstract: While the LMS algorithm and its normalized version (NLMS), have been thoroughly used and studied. Connections between the Kalman filter and the RLS algorithm have been established however, the connection between the Kalman filter and the LMS algorithm has not received much attention. By linking these two algorithms, a new normalized Kalman based LMS (KLMS) algorithm can be derived that has some advantages to the classical one. Their stability is guaranteed since they are a special case of the Kalman filter. More, they suggests a new way to control the step size, that results in good convergence properties for a large range of input signal powers, that occur in many applications. They prevent high measurement noise sensitivity that may occur in the NLMS algorithm for low order filters, like the ones used in OFDM equalization systems. In these paper, different algorithms based on the correlation form, information form and simplified versions of these are presented. The simplified form maintain the good convergence properties of the KLMS with slightly lower computational complexity.

A Prototype System for Selective Dissemination of Broadcast News in European Portuguese

Abstract: This paper describes ongoing work on selective dissemination of broadcast news. Our pipeline system includes several modules: audio preprocessing, speech recognition, and topic segmentation and indexation. The main goal of this work is to study the impact of earlier errors in the last modules. The impact of audio preprocessing errors is quite small on the speech recognition module, but quite significant in terms of topic segmentation. On the other hand, the impact of speech recognition errors on the topic segmentation and indexation modules is almost negligible. The diagnostic of the errors in these modules is a very important step for the improvement of the prototype of a media watch system described in this paper.

Adaptive motion estimation processor for autonomous video devices

Abstract: Motion estimation is the most demanding operation of a video encoder, corresponding to at least 80% of the overall computational cost. As a consequence, with the proliferation of autonomous and portable handheld devices that support digital video coding, data-adaptive motion estimation algorithms have been required to dynamically configure the search pattern not only to avoid unnecessary computations and memory accesses but also to save energy. This paper proposes an application-specific instruction set processor (ASIP) to implement data-adaptive motion estimation algorithms that is characterized by a specialized datapath and a minimum and optimized instruction set. Due to its low-power nature, this architecture is highly suitable to develop motion estimators for portable, mobile, and battery-supplied devices. Based on the proposed architecture and the considered adaptive algorithms, several motion estimators were synthesized both for a Virtex-II Pro XC2VP30 FPGA from Xilinx, integrated within an ML310 development platform, and using a StdCell library based on a 0.18 µm CMOS process. Experimental results show that the proposed architecture is able to estimate motion vectors in real time for QCIF and CIF video sequences with a very low-power consumption. Moreover, it is also able to adapt the operation to the available energy level in runtime. By adjusting the search pattern and setting up a more convenient operating frequency, it can change the power consumption in the interval between 1.6mW and 15 mW.

An authoring tool for an emergent narrative storytelling system

Abstract: In this paper we present the initial conceptual design of an authoring tool for the emergent narrative agent architecture FAtiMA that powers the virtual bullying drama FearNot!. We explain that the process of authoring emergent narrative to a large part consists of designing a planning domain for a virtual character planner and explain the difficulties this task poses to the non-technical author. After reviewing existing authoring tools and evaluating them in terms of their applicability to FAtiMA, we introduce a novel concept of approaching the authoring task, in which the author is playing through example story lines that are used to gradually increase the knowledge and intelligence of a virtual character. This concept is extended by a mixed initiative feature, which allows the author to cooperate with the character planners while providing the example stories. Finally we concretize our idea and explain our intended implementation of it within the FearNot! Framework. We believe that our design, although being specified with a particular architecture (FAtiMA) in mind, may provide some interesting ideas to others, who are trying to solve the authoring problem for interactive storytelling sys-

Design and implementation of a stream-based distributedcomputing platform using graphics processing units

Abstract: Anonymous use of computing resources spread over the world becomes one of the main goals in GRID environments. In GRID-based computing, the security of users or of contributors of computing resources is crucial to execute processes in a safe way. This paper proposes a new method for streambased processing in a distributed environment and also a novel method to solve the security matter under this kind of processing. It also presents the design of the distributed computing platform developed for stream-based processing, including the description of the local and remote execution methods, which are collectively designated by Caravela platform. The proposed flow-model is mapped on the distributed processing resources, connected through a network, by using the Caravela platform. This platform has been developed by the authors of this paper specifically for making use of the Graphics Processing Units available in recent personal computers. The paper also illustrates application of the Caravela platform to different types of processing, namely scientific computing and image/video processing. The presented experimental results show that significant improvements can be achieved with the use of GPUs against the use of general purpose processors.

High Voltage Tolerant Level Shifters and DCVSL in Standard Low Voltage CMOS Technologies

Abstract: In this paper, high voltage (HV) tolerant level-shifters with combinational functionality are proposed based on differential cascode voltage switch logic (DCVSL). These level-shifters are tolerant to supply voltages higher than the process limit for individual CMOS transistors. The proposed HV DCVSL level shifters are particularly useful when it is mandatory to ensure a specific behavior during out of the normal mode periods (power up; power down; reset; etc.). These high voltage tolerant logic circuits were used in the power block of buck converter designed in a standard 3.3 V, 0.13 mum CMOS process, powered by an input voltage range from 2.7 V to 4.2 V.

A Data-Driven Approach for Pipelining Sequences of Data-Dependent Loops

Abstract: Many video and image/signal processing applications can be structured as sequences of data-dependent tasks using a consumer/producer communication paradigm and are therefore amenable to pipelined execution. This paper presents an execution technique to speed-up the overall execution of successive, data-dependent tasks on a reconfigurable architecture. The technique pipelines sequences of data-dependent tasks by overlapping their execution subject to data-dependences. It decouples the concurrent data-path and control units and uses a custom, application data-driven, fine-grained synchronization and buffering scheme. In addition, the execution scheme allows for out-of- order, but data-dependent producer-consumer pairs not allowed by previous data-driven pipelining approaches. The approach has been exploited in the context of a high-level compiler targeting FPGAs. The preliminary experimental results reveal noticeable performance improvements and buffer size reductions for a number of benchmarks over traditional approaches.

Data buffering optimization methods toward a uniform programming interface for gpu-based applications

Abstract: The massive computational power available in off-the shelf Graphics Processing Units (GPUs) can pave the way for its usage in general purpose applications. Current interfaces to program GPU operation are still oriented towards graphics processing. This paper is focused in disparities on those programming interfaces and proposes an extension to of the recently developed Caravela library that supports streambased computation. This extension implements effective methods to counterbalance the disparities and differences in graphics runtime environments. Experimental results show that these methods improve performance of GPU-based applications by more than 50% and demonstrate that the proposed extended interface can be an effective solution for generalpurpose programming on GPUs.

Identification of FRBR works within bibliographic databases: an experiment with UNIMARC and duplicate detection techniques

Abstract: Many experiments and studies have been conducted on the application of FRBR as an implementation model for bibliographic databases, in order to improve the services of resource discovery and transmit better perception of the information spaces represented in catalogues One of these applications is the attempt to identify the FRBR work instances shared by several bibliographic records In our work we evaluate the applicability to this problem of techniques based on string similarity, used in duplicate detection procedures mainly by the database research community We describe the particularities of the application of these techniques to bibliographic data, and empirically compare the results obtained with these techniques to those obtained by current techniques, which are based on exact matching Experiments performed on the Portuguese national union catalogue show a significant improvement over currently used approaches

Extractive summarization of broadcast news: comparing strategies for European portuguese

Abstract: This paper presents the comparison between three methods for extractive summarization of Portuguese broadcast news: feature-based, Maximal Marginal Relevance, and Latent Semantic Analysis. The main goal is to understand the level of agreement among the automatic summaries and how they compare to summaries produced by non-professional human summarizers. Results were evaluated using the ROUGE-L metric. Maximal Marginal Relevance performed close to human summarizers. Both feature-based and Latent Semantic Analysis automatic summarizers performed close to each other and worse than Maximal Marginal Relevance, when compared to the summaries done by the human summarizers.