Abstract: Traffic control has been one of the most common and irritating problems since the time automobiles have hit the roads. Problems like traffic congestion have led to a significant time burden around the world and one significant solution to these problems can be the proper implementation of the Intelligent Transport System (ITS). It involves the integration of various tools like smart sensors, artificial intelligence, position technologies and mobile data services to manage traffic flow, reduce congestion and enhance driver's ability to avoid accidents during adverse weather. Road and traffic signs’ recognition is an emerging field of research in ITS. Classification problem of traffic signs needs to be solved as it is a major step in our journey towards building semi-autonomous/autonomous driving systems. The purpose of this work focuses on implementing an approach to solve the problem of traffic sign classification by developing a Convolutional Neural Network (CNN) classifier using the GTSRB (German Traffic Sign Recognition Benchmark) dataset. Rather than using hand-crafted features, our model addresses the concern of exploding huge parameters and data method augmentations. Our model achieved an accuracy of around 97.6% which is comparable to various state-of-the-art architectures.
Abstract: Environmental changes and major natural disasters are
most prevalent in the world due to the damage that humanity has
caused to nature and these damages directly affect the lives of
animals. Thus, the study of animal behavior and their interactions
with the environment can provide knowledge that guides researchers
and public agencies in preservation and conservation actions.
Exploratory analysis of animal movement can determine the patterns
of animal behavior and with technological advances the ability of
animals to be tracked and, consequently, behavioral studies have
been expanded. There is a lot of research on animal movement and
behavior, but we note that a proposal that combines resources and
allows for exploratory analysis of animal movement and provide
statistical measures on individual animal behavior and its interaction
with the environment is missing. The contribution of this paper is
to present the framework AniMoveMineR, a unified solution that
aggregates trajectory analysis and data mining techniques to explore
animal movement data and provide a first step in responding questions
about the animal individual behavior and their interactions with other
animals over time and space. We evaluated the framework through the
use of monitored jaguar data in the city of Miranda Pantanal, Brazil,
in order to verify if the use of AniMoveMineR allows to identify the
interaction level between these jaguars. The results were positive and
provided indications about the individual behavior of jaguars and
about which jaguars have the highest or lowest correlation.
Abstract: This paper presents and benchmarks a number of
end-to-end Deep Learning based models for metaphor detection in
Greek. We combine Convolutional Neural Networks and Recurrent
Neural Networks with representation learning to bear on the metaphor
detection problem for the Greek language. The models presented
achieve exceptional accuracy scores, significantly improving the
previous state-of-the-art results, which had already achieved accuracy
0.82. Furthermore, no special preprocessing, feature engineering or
linguistic knowledge is used in this work. The methods presented
achieve accuracy of 0.92 and F-score 0.92 with Convolutional
Neural Networks (CNNs) and bidirectional Long Short Term Memory
networks (LSTMs). Comparable results of 0.91 accuracy and 0.91
F-score are also achieved with bidirectional Gated Recurrent Units
(GRUs) and Convolutional Recurrent Neural Nets (CRNNs). The
models are trained and evaluated only on the basis of training tuples,
the related sentences and their labels. The outcome is a state-of-the-art
collection of metaphor detection models, trained on limited labelled
resources, which can be extended to other languages and similar
tasks.
Abstract: Machine learning (ML) can be implemented in Wireless Sensor Networks (WSNs) as a central solution or distributed solution where the ML is embedded in the nodes. Embedding improves privacy and may reduce prediction delay. In addition, the number of transmissions is reduced. However, quality factors such as prediction accuracy, fault detection efficiency and coordinated control of the overall system suffer. Here, we discuss and highlight the trade-offs that should be considered when choosing between embedding and centralized ML, especially for multihop networks. In addition, we present estimations that demonstrate the energy trade-offs between embedded and centralized ML. Although the total network energy consumption is lower with central prediction, it makes the network more prone for partitioning due to the high forwarding load on the one-hop nodes. Moreover, the continuous improvements in the number of operations per joule for embedded devices will move the energy balance toward embedded prediction.
Abstract: Multimodal image registration is a profoundly complex
task which is why deep learning has been used widely to address it in
recent years. However, two main challenges remain: Firstly, the lack
of ground truth data calls for an unsupervised learning approach,
which leads to the second challenge of defining a feasible loss
function that can compare two images of different modalities to judge
their level of alignment. To avoid this issue altogether we implement a
generative adversarial network consisting of two registration networks
GAB, GBA and two discrimination networks DA, DB connected by
spatial transformation layers. GAB learns to generate a deformation
field which registers an image of the modality B to an image of the
modality A. To do that, it uses the feedback of the discriminator DB
which is learning to judge the quality of alignment of the registered
image B. GBA and DA learn a mapping from modality A to modality
B. Additionally, a cycle-consistency loss is implemented. For this,
both registration networks are employed twice, therefore resulting in
images ˆA, ˆB which were registered to ˜B, ˜A which were registered
to the initial image pair A, B. Thus the resulting and initial images
of the same modality can be easily compared. A dataset of liver
CT and MRI was used to evaluate the quality of our approach and
to compare it against learning and non-learning based registration
algorithms. Our approach leads to dice scores of up to 0.80 ± 0.01
and is therefore comparable to and slightly more successful than
algorithms like SimpleElastix and VoxelMorph.
Abstract: Counterfeit goods and documents are a global problem, which needs more and more sophisticated methods of resolving it. Existing techniques using watermarking or embedding symbols on objects are not suitable for all use cases. To address those special needs, we created complete system allowing authentication of paper documents and physical objects with flat surface. Objects are marked using orientation independent and resistant to camera noise 2D graphic codes, named DotAuth. Based on the identifier stored in 2D code, the system is able to perform basic authentication and allows to conduct more sophisticated analysis methods, e.g., relying on augmented reality and physical properties of the object. In this paper, we present the complete architecture, algorithms and applications of the proposed system. Results of the features comparison of the proposed solution and other products are presented as well, pointing to the existence of many advantages that increase usability and efficiency in the means of protecting physical objects.
Abstract: This paper discusses the idea of capturing an expert’s knowledge in the form of human understandable rules and then inserting these rules into a dynamic cell structure (DCS) neural network. The DCS is a form of self-organizing map that can be used for many purposes, including classification and prediction. This particular neural network is considered to be a topology preserving network that starts with no pre-structure, but assumes a structure once trained. The DCS has been used in mission and safety-critical applications, including adaptive flight control and health-monitoring in aerial vehicles. The approach is to insert expert knowledge into the DCS before training. Rules are translated into a pre-structure and then training data are presented. This idea has been demonstrated using the well-known Iris data set and it has been shown that inserting the pre-structure results in better accuracy with the same training.
Abstract: Thermoelectric technology is currently being used in many industrial applications for cooling, heating and generating electricity. This research mainly focuses on using thermoelectric to cool down high-speed computer chips at different operating conditions. A previously developed and validated three-dimensional model for optimizing and assessing the performance of cascaded thermoelectric and non-cascaded thermoelectric is used in this study to investigate the possibility of decreasing the hotspot temperature of computer chip. Additionally, a test assembly is built and tested at steady-state and transient conditions. The obtained optimum thermoelectric current at steady-state condition is used to conduct a number of pulsed tests (i.e. transient tests) with different shapes to cool the computer chips hotspots. The results of the steady-state tests showed that at hotspot heat rate of 15.58 W (5.97 W/cm2), using thermoelectric current of 4.5 A has resulted in decreasing the hotspot temperature at open circuit condition (89.3 °C) by 50.1 °C. Maximum and minimum hotspot temperatures have been affected by ON and OFF duration of the electrical current pulse. Maximum hotspot temperature was resulted by longer OFF pulse period. In addition, longer ON pulse period has generated the minimum hotspot temperature.
Abstract: In recent years, the number of document data has been
increasing since the spread of the Internet. Many methods have been
studied for extracting topics from large document data. We proposed
Independent Topic Analysis (ITA) to extract topics independent of
each other from large document data such as newspaper data. ITA is a
method for extracting the independent topics from the document data
by using the Independent Component Analysis. The topic represented
by ITA is represented by a set of words. However, the set of words
is quite different from the topics the user imagines. For example,
the top five words with high independence of a topic are as follows.
Topic1 = {"scor", "game", "lead", "quarter", "rebound"}. This Topic
1 is considered to represent the topic of "SPORTS". This topic name
"SPORTS" has to be attached by the user. ITA cannot name topics.
Therefore, in this research, we propose a method to obtain topics easy
for people to understand by using the web search engine, topics given
by the set of words given by independent topic analysis. In particular,
we search a set of topical words, and the title of the homepage of
the search result is taken as the topic name. And we also use the
proposed method for some data and verify its effectiveness.
Abstract: To claim the ownership for an executable program is a non-trivial task. An emerging direction is to add a watermark to the program such that the watermarked program preserves the original program’s functionality and removing the watermark would heavily destroy the functionality of the watermarked program. In this paper, the first watermarking signature scheme with the watermark and the constraint function hidden in the symmetric key setting is constructed. The scheme uses well-known techniques of lattice trapdoors and a lattice evaluation. The watermarking signature scheme is unforgeable under the Short Integer Solution (SIS) assumption and satisfies other security requirements such as the unremovability security property.
Abstract: This paper compares the multipath mitigation
performance of code correlation reference waveform receivers
with variable and fixed window width, for binary offset carrier
and multiplexed binary offset carrier signals typically used in
global navigation satellite systems. In the variable window width
method, such width is iteratively reduced until the distortion on
the discriminator with multipath is eliminated. This distortion is
measured as the Euclidean distance between the actual discriminator
(obtained with the incoming signal), and the local discriminator
(generated with a local copy of the signal). The variable window
width have shown better performance compared to the fixed window
width. In particular, the former yields zero error for all delays for
the BOC and MBOC signals considered, while the latter gives
rather large nonzero errors for small delays in all cases. Due to
its computational simplicity, the variable window width method is
perfectly suitable for implementation in low-cost receivers.
Abstract: This research presents the first constant approximation
algorithm to the p-median network design problem with multiple
cable types. This problem was addressed with a single cable type and
there is a bifactor approximation algorithm for the problem. To the
best of our knowledge, the algorithm proposed in this paper is the first
constant approximation algorithm for the p-median network design
with multiple cable types. The addressed problem is a combination of
two well studied problems which are p-median problem and network
design problem. The introduced algorithm is a random sampling
approximation algorithm of constant factor which is conceived by
using some random sampling techniques form the literature. It is
based on a redistribution Lemma from the literature and a steiner tree
problem as a subproblem. This algorithm is simple, and it relies on the
notions of random sampling and probability. The proposed approach
gives an approximation solution with one constant ratio without
violating any of the constraints, in contrast to the one proposed in the
literature. This paper provides a (21 + 2)-approximation algorithm
for the p-median network design problem with multiple cable types
using random sampling techniques.
Abstract: With the recent advance of the deep neural network, we observe new applications of NLP (natural language processing) and CV (computer vision) powered by deep neural networks for processing business documents. However, creating a real-world document processing system needs to integrate several NLP and CV tasks, rather than treating them separately. There is a need to have a unified approach for processing documents containing textual and graphical elements with rich formats, diverse layout arrangement, and distinct semantics. In this paper, a framework that fulfills this unified approach is presented. The framework includes a representation model definition for holding the information generated by various tasks and specifications defining the coordination between these tasks. The framework is a blueprint for building a system that can process documents with rich formats, styles, and multiple types of elements. The flexible and lightweight design of the framework can help build a system for diverse business scenarios, such as contract monitoring and reviewing.
Abstract: Due to the development of the current civilization, one must create suitable models of its pervasive massive phenomena. Such a phenomenon is the digital transformation, which has a substantial number of disciplined, methodical interpretations forming the diversified reflection. This reflection could be understood pragmatically as the current temporal, a local differential state of knowledge. The model of the discursive space is proposed as a model for the analysis and description of this knowledge. Discursive space is understood as an autonomous multidimensional space where separate discourses traverse specific trajectories of what can be presented in multidimensional parallel coordinate system. Discursive space built on the world of facts preserves the complex character of that world. Digital transformation as a discursive space has a relativistic character that means that at the same time, it is created by the dynamic discourses and these discourses are molded by the shape of this space.
Abstract: In sports, individuals and teams are typically interested
in final rankings. Final results, such as times or distances, dictate
these rankings, also known as places. Places can be further associated
with ordered random variables, commonly referred to as order
statistics. In this work, we introduce a simple, yet accurate order
statistical ordinal regression function that predicts relay race places
with changeover-times. We call this function the Fenton-Wilkinson
Order Statistics model. This model is built on the following educated
assumption: individual leg-times follow log-normal distributions.
Moreover, our key idea is to utilize Fenton-Wilkinson approximations
of changeover-times alongside an estimator for the total number
of teams as in the notorious German tank problem. This original
place regression function is sigmoidal and thus correctly predicts
the existence of a small number of elite teams that significantly
outperform the rest of the teams. Our model also describes how place
increases linearly with changeover-time at the inflection point of the
log-normal distribution function. With real-world data from Jukola
2019, a massive orienteering relay race, the model is shown to be
highly accurate even when the size of the training set is only 5%
of the whole data set. Numerical results also show that our model
exhibits smaller place prediction root-mean-square-errors than linear
regression, mord regression and Gaussian process regression.
Abstract: In linear estimation, the traditional Kalman filter uses the Kalman filter gain in order to produce estimation and prediction of the n-dimensional state vector using the m-dimensional measurement vector. The computation of the Kalman filter gain requires the inversion of an m x m matrix in every iteration. In this paper, a variation of the Kalman filter eliminating the Kalman filter gain is proposed. In the time varying case, the elimination of the Kalman filter gain requires the inversion of an n x n matrix and the inversion of an m x m matrix in every iteration. In the time invariant case, the elimination of the Kalman filter gain requires the inversion of an n x n matrix in every iteration. The proposed Kalman filter gain elimination algorithm may be faster than the conventional Kalman filter, depending on the model dimensions.
Abstract: Program synthesis is the task to automatically generate
programs based on user specification. In this paper, we present a
framework that synthesizes programs from flow charts that serve as
accurate and intuitive specification. In order doing so, we propose a
deep neural network called GRCNN that recognizes graph structure
from its image. GRCNN is trained end-to-end, which can predict edge
and node information of the flow chart simultaneously. Experiments
show that the accuracy rate to synthesize a program is 66.4%, and
the accuracy rates to recognize edge and node are 94.1% and 67.9%,
respectively. On average, it takes about 60 milliseconds to synthesize
a program.
Abstract: Nowadays, the abuse of keyloggers is one of the most widespread approaches to steal sensitive information. In this paper, we propose an On-Screen Prompts Approach to Keyloggers (OSPAK) and its analysis, which is installed in public computers. OSPAK utilizes a canvas to cue users when their keystrokes are going to be logged or ignored by OSPAK. This approach can protect computers against recoding sensitive inputs, which obfuscates keyloggers with letters inserted among users' keystrokes. It adds a canvas below each password field in a webpage and consists of three parts: two background areas, a hit area and a moving foreground object. Letters at different valid time intervals are combined in accordance with their time interval orders, and valid time intervals are interleaved with invalid time intervals. It utilizes animation to visualize valid time intervals and invalid time intervals, which can be integrated in a webpage as a browser extension. We have tested it against a series of known keyloggers and also performed a study with 95 users to evaluate how easily the tool is used. Experimental results made by volunteers show that OSPAK is a simple approach.
Abstract: The evolution in mobile broadband technologies has allowed to increase the download rates in users considering the current services. The evaluation of technical parameters at the link level is of vital importance to validate the quality and veracity of the connection, thus avoiding large losses of data, time and productivity. Some of these failures may occur between the eNodeB (Evolved Node B) and the user equipment (UE), so the link between the end device and the base station can be observed. LTE (Long Term Evolution) is considered one of the IP-oriented mobile broadband technologies that work stably for data and VoIP (Voice Over IP) for those devices that have that feature. This research presents a technical analysis of the connection and channeling processes between UE and eNodeB with the TAC (Tracking Area Code) variables, and analysis of performance variables (Throughput, Signal to Interference and Noise Ratio (SINR)). Three measurement scenarios were proposed in the city of Bogotá using QualiPoc, where two operators were evaluated (Operator 1 and Operator 2). Once the data were obtained, an analysis of the variables was performed determining that the data obtained in transmission modes vary depending on the parameters BLER (Block Error Rate), performance and SNR (Signal-to-Noise Ratio). In the case of both operators, differences in transmission modes are detected and this is reflected in the quality of the signal. In addition, due to the fact that both operators work in different frequencies, it can be seen that Operator 1, despite having spectrum in Band 7 (2600 MHz), together with Operator 2, is reassigning to another frequency, a lower band, which is AWS (1700 MHz), but the difference in signal quality with respect to the establishment with data by the provider Operator 2 and the difference found in the transmission modes determined by the eNodeB in Operator 1 is remarkable.
Abstract: This paper presents a method for single image dehazing based on dark channel prior (DCP). The property that the intensity of the dark channel gives an approximate thickness of the haze is used to estimate the transmission and atmospheric light. Instead of constant atmospheric light, the proposed method employs scene depth to estimate spatially varying atmospheric light as it truly occurs in nature. Haze imaging model together with the soft matting method has been used in this work to produce high quality haze free image. Experimental results demonstrate that the proposed approach produces better results than the classic DCP approach as color fidelity and contrast of haze free image are improved and no over-saturation in the sky region is observed. Further, lifting Haar wavelet transform is employed to reduce overall execution time by a factor of two to three as compared to the conventional approach.