Abstract: Most neural network (NN) models of human category learning use a gradient-based learning method, which assumes that locally-optimal changes are made to model parameters on each learning trial. This method tends to under predict variability in individual-level cognitive processes. In addition many recent models of human category learning have been criticized for not being able to replicate rapid changes in categorization accuracy and attention processes observed in empirical studies. In this paper we introduce stochastic learning algorithms for NN models of human category learning and show that use of the algorithms can result in (a) rapid changes in accuracy and attention allocation, and (b) different learning trajectories and more realistic variability at the individual-level.
Abstract: Message Passing Interface is widely used for Parallel
and Distributed Computing. MPICH and LAM are popular open
source MPIs available to the parallel computing community also
there are commercial MPIs, which performs better than MPICH etc.
In this paper, we discuss a commercial Message Passing Interface, CMPI
(C-DAC Message Passing Interface). C-MPI is an optimized
MPI for CLUMPS. It is found to be faster and more robust compared
to MPICH. We have compared performance of C-MPI and MPICH
on Gigabit Ethernet network.
Abstract: There have been many games developing simulation
of soccer games. Many of these games have been designed with
highly realistic features to attract more users. Many have also
incorporated better artificial intelligent (AI) similar to that in a real
soccer game. One of the challenging issues in a soccer game is the
cooperation, coordination and negotiation among distributed agents
in a multi-agent system. This paper focuses on the incorporation of
multi-agent technique in a soccer game domain. The better the
cooperation of a multi-agent team, the more intelligent the game will
be. Thus, past studies were done on the robotic soccer game because
of the better multi-agent system implementation. From this study, a
better approach and technique of multi-agent behavior could be
select to improve the author-s 2D online soccer game.
Abstract: Detecting object in video sequence is a challenging
mission for identifying, tracking moving objects. Background
removal considered as a basic step in detected moving objects tasks.
Dual static cameras placed in front and rear moving platform
gathered information which is used to detect objects. Background
change regarding with speed and direction moving platform, so
moving objects distinguished become complicated. In this paper, we
propose framework allows detection moving object with variety of
speed and direction dynamically. Object detection technique built on
two levels the first level apply background removal and edge
detection to generate moving areas. The second level apply Moving
Areas Filter (MAF) then calculate Correlation Score (CS) for
adjusted moving area. Merging moving areas with closer CS and
marked as moving object. Experiment result is prepared on real scene
acquired by dual static cameras without overlap in sense. Results
showing accuracy in detecting objects compared with optical flow
and Mixture Module Gaussian (MMG), Accurate ratio produced to
measure accurate detection moving object.
Abstract: Data mining and knowledge engineering have become a tough task due to the availability of large amount of data in the web nowadays. Validity and reliability of data also become a main debate in knowledge acquisition. Besides, acquiring knowledge from different languages has become another concern. There are many language translators and corpora developed but the function of these translators and corpora are usually limited to certain languages and domains. Furthermore, search results from engines with traditional 'keyword' approach are no longer satisfying. More intelligent knowledge engineering agents are needed. To address to these problems, a system known as Multilingual Word Semantic Network is proposed. This system adapted semantic network to organize words according to concepts and relations. The system also uses open source as the development philosophy to enable the native language speakers and experts to contribute their knowledge to the system. The contributed words are then defined and linked using lexical and semantic relations. Thus, related words and derivatives can be identified and linked. From the outcome of the system implementation, it contributes to the development of semantic web and knowledge engineering.
Abstract: Considering the theory of attribute grammars, we use
logical formulas instead of traditional functional semantic rules.
Following the decoration of a derivation tree, a suitable algorithm
should maintain the consistency of the formulas together with the
evaluation of the attributes. This may be a Prolog-like resolution, but
this paper examines a somewhat different strategy, based on
production specialization, local consistency and propagation: given a
derivation tree, it is interactively decorated, i.e. incrementally
checked and evaluated. The non-directed dependencies are
dynamically directed during attribute evaluation.
Abstract: This study proposes novel hybrid social network analysis and collaborative filtering approach to enhance the performance of recommender systems. The proposed model selects subgroups of users in Internet community through social network analysis (SNA), and then performs clustering analysis using the information about subgroups. Finally, it makes recommendations using cluster-indexing CF based on the clustering results. This study tries to use the cores in subgroups as an initial seed for a conventional clustering algorithm. This model chooses five cores which have the highest value of degree centrality from SNA, and then performs clustering analysis by using the cores as initial centroids (cluster centers). Then, the model amplifies the impact of friends in social network in the process of cluster-indexing CF.
Abstract: The Tropical Data Hub (TDH) is a virtual research environment that provides researchers with an e-research infrastructure to congregate significant tropical data sets for data reuse, integration, searching, and correlation. However, researchers often require data and metadata synthesis across disciplines for crossdomain analyses and knowledge discovery. A triplestore offers a semantic layer to achieve a more intelligent method of search to support the synthesis requirements by automating latent linkages in the data and metadata. Presently, the benchmarks to aid the decision of which triplestore is best suited for use in an application environment like the TDH are limited to performance. This paper describes a new evaluation tool developed to analyze both features and performance. The tool comprises a weighted decision matrix to evaluate the interoperability, functionality, performance, and support availability of a range of integrated and native triplestores to rank them according to requirements of the TDH.
Abstract: Needs of an efficient information retrieval in recent
years in increased more then ever because of the frequent use of
digital information in our life. We see a lot of work in the area of
textual information but in multimedia information, we cannot find
much progress. In text based information, new technology of data
mining and data marts are now in working that were started from the
basic concept of database some where in 1960.
In image search and especially in image identification,
computerized system at very initial stages. Even in the area of image
search we cannot see much progress as in the case of text based
search techniques. One main reason for this is the wide spread roots
of image search where many area like artificial intelligence,
statistics, image processing, pattern recognition play their role. Even
human psychology and perception and cultural diversity also have
their share for the design of a good and efficient image recognition
and retrieval system.
A new object based search technique is presented in this paper
where object in the image are identified on the basis of their
geometrical shapes and other features like color and texture where
object-co-relation augments this search process.
To be more focused on objects identification, simple images are
selected for the work to reduce the role of segmentation in overall
process however same technique can also be applied for other
images.
Abstract: The paper presents a multimodal approach for biometric authentication, based on multiple classifiers. The proposed solution uses a post-classification biometric fusion method in which the biometric data classifiers outputs are combined in order to improve the overall biometric system performance by decreasing the classification error rates. The paper shows also the biometric recognition task improvement by means of a carefully feature selection, as much as not all of the feature vectors components support the accuracy improvement.
Abstract: In this paper we introduce new data oriented modeling
of uniform random variable well-matched with computing systems. Due to this conformity with current computers structure, this modeling will be efficiently used in statistical inference.
Abstract: Realistic 3D face model is more precise in representing
pose, illumination, and expression of face than 2D face model so that it
can be utilized usefully in various applications such as face recognition,
games, avatars, animations, and etc.
In this paper, we propose a 3D face modeling method based on 3D
dense morphable shape model. The proposed 3D modeling method
first constructs a 3D dense morphable shape model from 3D face scan
data obtained using a 3D scanner. Next, the proposed method extracts
and matches facial landmarks from 2D image sequence containing a
face to be modeled, and then reconstructs 3D vertices coordinates of
the landmarks using a factorization-based SfM technique. Then, the
proposed method obtains a 3D dense shape model of the face to be
modeled by fitting the constructed 3D dense morphable shape model
into the reconstructed 3D vertices. Also, the proposed method makes a
cylindrical texture map using 2D face image sequence. Finally, the
proposed method generates a 3D face model by rendering the 3D dense
face shape model using the cylindrical texture map. Through building
processes of 3D face model by the proposed method, it is shown that
the proposed method is relatively easy, fast and precise.
Abstract: We present an implementation of an Online Exhibition System (OES) web service(s) that reflects our experiences with using web service development packages and software process models. The system provides major functionality that exists in similar packages. While developing such a complex web service, we gained insightful experience (i) in the traditional software development processes: waterfall model and evolutionary development and their fitness to web services development, (ii) in the fitness and effectiveness of a major web services development kit.
Abstract: On-line handwritten scripts are usually dealt with pen tip traces from pen-down to pen-up positions. Time evaluation of the pen coordinates is also considered along with trajectory information. However, the data obtained needs a lot of preprocessing including filtering, smoothing, slant removing and size normalization before recognition process. Instead of doing such lengthy preprocessing, this paper presents a simple approach to extract the useful character information. This work evaluates the use of the counter- propagation neural network (CPN) and presents feature extraction mechanism in full detail to work with on-line handwriting recognition. The obtained recognition rates were 60% to 94% using the CPN for different sets of character samples. This paper also describes a performance study in which a recognition mechanism with multiple thresholds is evaluated for counter-propagation architecture. The results indicate that the application of multiple thresholds has significant effect on recognition mechanism. The method is applicable for off-line character recognition as well. The technique is tested for upper-case English alphabets for a number of different styles from different peoples.
Abstract: This study analyzes the effect of discretization on
classification of datasets including continuous valued features. Six
datasets from UCI which containing continuous valued features are
discretized with entropy-based discretization method. The
performance improvement between the dataset with original features
and the dataset with discretized features is compared with k-nearest
neighbors, Naive Bayes, C4.5 and CN2 data mining classification
algorithms. As the result the classification accuracies of the six
datasets are improved averagely by 1.71% to 12.31%.
Abstract: This paper describes a new approach of classification
using genetic programming. The proposed technique consists of
genetically coevolving a population of non-linear transformations on
the input data to be classified, and map them to a new space with a
reduced dimension, in order to get a maximum inter-classes
discrimination. The classification of new samples is then performed
on the transformed data, and so become much easier. Contrary to the
existing GP-classification techniques, the proposed one use a
dynamic repartition of the transformed data in separated intervals, the
efficacy of a given intervals repartition is handled by the fitness
criterion, with a maximum classes discrimination. Experiments were
first performed using the Fisher-s Iris dataset, and then, the KDD-99
Cup dataset was used to study the intrusion detection and
classification problem. Obtained results demonstrate that the
proposed genetic approach outperform the existing GP-classification
methods [1],[2] and [3], and give a very accepted results compared to
other existing techniques proposed in [4],[5],[6],[7] and [8].
Abstract: This research investigates the design of a low-cost 3D
spatial interaction approach using the Wii Remote for immersive
Head-Mounted Display (HMD) virtual reality. Current virtual reality
applications that incorporate the Wii Remote are either desktop
virtual reality applications or systems that use large screen displays.
However, the requirements for an HMD virtual reality system differ
from such systems. This is mainly because in HMD virtual reality,
the display screen does not remain at a fixed location. The user views
the virtual environment through display screens that are in front of
the user-s eyes and when the user moves his/her head, these screens
move as well. This means that the display has to be updated in realtime
based on where the user is currently looking. Normal usage of
the Wii Remote requires the controller to be pointed in a certain
direction, typically towards the display. This is too restrictive for
HMD virtual reality systems that ideally require the user to be able to
turn around in the virtual environment. Previous work proposed a
design to achieve this, however it suffered from a number of
drawbacks. The aim of this study is to look into a suitable method of
using the Wii Remote for 3D interaction in a space around the user
for HMD virtual reality. This paper presents an overview of issues
that had to be considered, the system design as well as experimental
results.
Abstract: The innovative intelligent fuzzy weighted input
estimation method (FWIEM) can be applied to the inverse heat
transfer conduction problem (IHCP) to estimate the unknown
time-varying heat flux of the multilayer materials as presented in this
paper. The feasibility of this method can be verified by adopting the
temperature measurement experiment. The experiment modular may
be designed by using the copper sample which is stacked up 4
aluminum samples with different thicknesses. Furthermore, the
bottoms of copper samples are heated by applying the standard heat
source, and the temperatures on the tops of aluminum are measured by
using the thermocouples. The temperature measurements are then
regarded as the inputs into the presented method to estimate the heat
flux in the bottoms of copper samples. The influence on the estimation
caused by the temperature measurement of the sample with different
thickness, the processing noise covariance Q, the weighting factor γ ,
the sampling time interval Δt , and the space discrete interval Δx ,
will be investigated by utilizing the experiment verification. The
results show that this method is efficient and robust to estimate the
unknown time-varying heat input of the multilayer materials.
Abstract: Nowadays, the rapid development of multimedia
and internet allows for wide distribution of digital media data.
It becomes much easier to edit, modify and duplicate digital
information Besides that, digital documents are also easy to
copy and distribute, therefore it will be faced by many
threatens. It-s a big security and privacy issue with the large
flood of information and the development of the digital
format, it become necessary to find appropriate protection
because of the significance, accuracy and sensitivity of the
information. Nowadays protection system classified with more
specific as hiding information, encryption information, and
combination between hiding and encryption to increase information
security, the strength of the information hiding science is due to the
non-existence of standard algorithms to be used in hiding secret
messages. Also there is randomness in hiding methods such as
combining several media (covers) with different methods to pass a
secret message. In addition, there are no formal methods to be
followed to discover the hidden data. For this reason, the task of this
research becomes difficult. In this paper, a new system of information
hiding is presented. The proposed system aim to hidden information
(data file) in any execution file (EXE) and to detect the hidden file
and we will see implementation of steganography system which
embeds information in an execution file. (EXE) files have been
investigated. The system tries to find a solution to the size of the
cover file and making it undetectable by anti-virus software. The
system includes two main functions; first is the hiding of the
information in a Portable Executable File (EXE), through the
execution of four process (specify the cover file, specify the
information file, encryption of the information, and hiding the
information) and the second function is the extraction of the hiding
information through three process (specify the steno file, extract the
information, and decryption of the information). The system has
achieved the main goals, such as make the relation of the size of the
cover file and the size of information independent and the result file
does not make any conflict with anti-virus software.
Abstract: Software crisis refers to the situation in which the developers are not able to complete the projects within time and budget constraints and moreover these overscheduled and over budget projects are of low quality as well. Several methodologies have been adopted form time to time to overcome this situation and now in the focus is component based software engineering. In this approach, emphasis is on reuse of already existing software artifacts. But the results can not be achieved just by preaching the principles; they need to be practiced as well. This paper highlights some of the very basic elements of this approach, which has to be in place to get the desired goals of high quality, low cost with shorter time-to-market software products.