9 results
Search Results
Now showing 1 - 9 of 9
Article A Wavelet-Based Feature Set for Recognizing Pulse Repetition Interval Modulation Patterns(2016) Gençol, Kenan; At, Nuray; Kara, AlıThis paper presents a new feature set for the problem of recognizing pulse repetition interval (PRI) modulation patterns. The recognition is based upon the features extracted from the multiresolution decomposition of different types of PRI modulated sequences. Special emphasis is placed on the recognition of jittered and stagger type PRI sequences due to the fact that these types of PRI sequences appear predominantly in modern electronic warfare environments for some specific mission requirements and recognition of them is heavily based on histogram features. We test our method with a broad range of PRI modulation parameters. Simulation results show that the proposed feature set is highly robust and separates jittered, stagger, and other modulation patterns very well. Especially for the stagger type of PRI sequences, wavelet-based features outperform conventional histogram-based features. Advantages of the proposed feature set along with its robustness criteria are analyzed in detail.Article Quality of Service Assessment: a Case Study on Performance Benchmarking of Cellular Network Operators in Turkey(2015) Kadıoğlu, Rana; Dalveren, Yaser; Kara, AliAbstract: This paper presents findings on performance benchmarking of cellular network operators in Turkey. Bench- marking is based on measurements of standard key performance indicators (KPIs) in one of the metropolitan cities of Turkey, Ankara. Performance benchmarking is formulated by incorporating customer perception by conducting surveys on how important KPIs are from the user s point of view. KPIs are measured, with standard test equipment, by drive test method on specified routes. According to the performance benchmarking results, the GSM and UMTS network operators achieving the best performance were determined in Ankara. Speech qualities of network operators, as the most popular service, were also evaluated by several statistical methods including pdf/cdf analysis and chi-square and Fisher s exact tests. The network operator providing the highest speech quality in Ankara was determined with the methods applied. Overall, the results and approaches on benchmarking of cellular networks in Turkey are reported for the first time in this paper. The approaches proposed in this paper could be adapted to wide-scale benchmarking of services in cellular networks.Article Citation - WoS: 12Citation - Scopus: 12A Polarity Calculation Approach for Lexicon-Based Turkish Sentiment Analysis(Tubitak Scientific & Technological Research Council Turkey, 2019) Yurtalan, Gökhan; Koyuncu, Murat; Turhan, ÇiğdemSentiment analysis attempts to resolve the senses or emotions that a writer or speaker intends to send across tothe people about an object or event. It generally uses natural language processing and/or artificial intelligence techniquesfor processing electronic documents and mining the opinion specified in the content. In recent years, researchers haveconducted many successful sentiment analysis studies for the English language which consider many words and wordgroups that set emotion polarities arising from the English grammar structure, and then use datasets to test theirperformance. However, there are only a limited number of studies for the Turkish language, and these studies have lowerperformance results compared to those studies for English. The reasons for this can be incorrect translation of datasetsfrom English into Turkish and ignoring the special grammar structures in the latter. In this study, special Turkish wordsand linguistic constructs which affect the polarity of a sentence are determined with the aid of a Turkish linguist, and anappropriate lexicon-based polarity determination and calculation approach is introduced for this language. The proposedmethodology is tested using different datasets collected from Twitter, and the test results show that the proposed systemachieves better accuracy than the previously developed lexical-based sentiment analysis systems for Turkish. The authorsconclude that especially analysis of word groups increases the overall performance of the system significantly.Article Fitting a Recurrent Dynamical Neural Network To Neural Spiking Data: Tackling the Sigmoidal Gain Function Issues(Tubitak Scientific & Technological Research Council Turkey, 2019) Doruk, Reşat ÖzgürThis is a continuation of a recent study (Doruk RO, Zhang K. Fitting of dynamic recurrent neural networkmodels to sensory stimulus-response data. J Biol Phys 2018; 44: 449-469), where a continuous time dynamical recurrentneural network is fitted to neural spiking data. In this research, we address the issues arising from the inclusion ofsigmoidal gain function parameters to the estimation algorithm. The neural spiking data will be obtained from the samemodel as that of Doruk and Zhang, but we propose a different model for identification. This will also be a continuoustime recurrent neural network, but with generic sigmoidal gains. The simulation framework and estimation algorithmsare kept similar to that of Doruk and Zhang so that we can have a solid base to compare the results. We evaluatethe estimation performance in two different ways. First, we compare the firing rate responses of the original and theestimated model. We find that responses of both models to the same stimuli are similar. Secondly, we evaluate variationsof the standard deviations of the estimates against a number of samples and stimulus parameters. They show a similarpattern to that of Doruk and Zhang. We thus conclude that our model serves as a reasonable alternative provided thatfiring rate is the response of interest (to any stimulus).Article Neuron Modeling: Estimating the Parameters of a Neuron Model From Neural Spiking Data(2018) Doruk, Reşat ÖzgürWe present a modeling study aiming at the estimation of the parameters of a single neuron model from neuralspiking data. The model receives a stimulus as input and provides the firing rate of the neuron as output. The neuralspiking data will be obtained from point process simulation. The resultant data will be used in parameter estimationbased on the inhomogeneous Poisson maximum likelihood method. The model will be stimulated by various forms ofstimuli, which are modeled by a Fourier series (FS), exponential functions, and radial basis functions (RBFs). Tabulatedresults presenting cases with different sample sizes (# of repeated trials), stimulus component sizes (FS and RBF),amplitudes, and frequency ranges (FS) will be presented to validate the approach and provide a means of comparison.The results showed that regardless of the stimulus type, the most effective parameter on the estimation performanceappears to be the sample size. In addition, the lowest variance of the estimates is obtained when a Fourier series stimulusis applied in the estimation.Article On the Independence of Statistical Randomness Tests Included in the Nist Test Suite(2017) Sulak, Fatih; Uğuz, Muhiddin; Koçak, Onur; Doğanaksoy, AliRandom numbers and random sequences are used to produce vital parts of cryptographic algorithms such as encryption keys and therefore the generation and evaluation of random sequences in terms of randomness are vital. Test suites consisting of a number of statistical randomness tests are used to detect the nonrandom characteristics of the sequences. Construction of a test suite is not an easy task. On one hand, the coverage of a suite should be wide; that is, it should compare the sequence under consideration from many different points of view with true random sequences. On the other hand, an overpopulated suite is expensive in terms of running time and computing power. Unfortunately, this trade-off is not addressed in detail in most of the suites in use. An efficient suite should avoid use of similar tests, while still containing sufficiently many. A single statistical test gives a measure for the randomness of the data. A collection of tests in a suite give a collection of measures. Obtaining a single value from this collection of measures is a difficult task and so far there is no conventional or strongly recommended method for this purpose. This work focuses on the evaluation of the randomness of data to give a uni ed result that considers all statistical information obtained from different tests in the suite. A natural starting point of research in this direction is to investigate correlations between test results and to study the independences of each from others. It is started with the concept of independence. As it is complicated enough to work even with one test function, theoretical investigation of dependence between many of them in terms of conditional probabilities is a much more difficult task. With this motivation, in this work it is tried to get some experimental results that may lead to theoretical results in future works. As experimental results may re ect properties of the data set under consideration, work is done on various types of large data sets hoping to get results that give clues about the theoretical results. For a collection of statistical randomness tests, the tests in the NIST test suite are considered. Tests in the NIST suite that can be applied to sequences shorter than 38,912 bits are analyzed. Based on the correlation of the tests at extreme values, the dependencies of the tests are found. Depending on the coverage of a test suite, a new concept, the coverage efficiency of a test suite, is de ned, and using this concept, the most efficient, the least efficient, and the optimal subsuites of the NIST suite are determined. Moreover, the marginal bene t of each test, which also helps one to understand the contribution of each individual test to the coverage efficiency of the NIST suite, is found. Furthermore, an efficient subsuite that contains ve statistical randomness tests is proposed.Article Selective Word Encoding for Effective Text Representation(Tubitak Scientific & Technological Research Council Turkey, 2019) Özkan, Savaş; Özkan, AkınDetermining the category of a text document from its semantic content is highly motivated in the literatureand it has been extensively studied in various applications. Also, the compact representation of the text is a fundamental step in achieving precise results for the applications and the studies are generously concentrated to improve itsperformance. In particular, the studies which exploit the aggregation of word-level representations are the mainstreamtechniques used in the problem. In this paper, we tackle text representation to achieve high performance in differenttext classification tasks. Throughout the paper, three critical contributions are presented. First, to encode the wordlevel representations for each text, we adapt a trainable orderless aggregation algorithm to obtain a more discriminativeabstract representation by transforming word vectors to the text-level representation. Second, we propose an effectiveterm-weighting scheme to compute the relative importance of words from the context based on their conjunction with theproblem in an end-to-end learning manner. Third, we present a weighted loss function to mitigate the class-imbalanceproblem between the categories. To evaluate the performance, we collect two distinct datasets as Turkish parliamentrecords (i.e. written speeches of four major political parties including 30731/7683 train and test documents) and newspaper articles (i.e. daily articles of the columnists including 16000/3200 train and test documents) whose data is availableon the web. From the results, the proposed method introduces significant performance improvements to the baselinetechniques (i.e. VLAD and Fisher Vector) and achieves 0.823% and 0.878% true prediction accuracies for the partymembership and the estimation of the category of articles respectively. The performance validates that the proposed contributions (i.e. trainable word-encoding model, trainable term-weighting scheme and weighted loss function) significantlyoutperform the baselines.Article An Rfid Based Indoor Tracking Method for Navigating Visually Impaired People(2010) Öktem, Ruşen; Aydın, ElifThis paper tackles the RFID based tracking problem in an obscured indoor environment. The proposed solution is an integral part of a navigation aid for guiding visually impaired people in a store. It uses RF signal strengths and is based on the Bayes Decision Theory. An observation vector is formed by received radio signal strength indication values, transmitted from three transmitters at distinct frequencies in the UHF band. The indoor area is divided into square grids, where each grid is considered as a class. The problem of tracking is expressed as classifying the observed radio signal strengths to the most likely class. A classification rule is formulated by incorporating a priori assumptions appropriate for the studied model. The proposed approach is tested in a laboratory environment. The results prove that the proposed approach is promising in tracking especially when the tracked person is guided by a system.Article Mutual Correlation of Nist Statistical Randomness Tests and Comparison of Theirsensitivities on Transformed Sequences(2017) Doğanaksoy, Ali; Sulak, Fatih; Uğuz, Muhiddin; Şeker, Okan; Akcengiz, ZiyaRandom sequences are widely used in many cryptographic applications and hence their generation is oneof the main research areas in cryptography. Statistical randomness tests are introduced to detect the weaknesses ornonrandom characteristics that a sequence under consideration may have. In the literature, there exist various statisticalrandomness tests and test suites, de ned as a collection of tests. An efficient test suite should consist of a number ofuncorrelated statistical tests each of which measures randomness from another point of view. `Being uncorrelated\\' is nota well-de ned or well-understood concept in the literature. In this work, we apply Pearson\\'s correlation test to measurethe correlation between the tests.In addition, we de ne ve new methods for transforming a sequence. Our motivation is to detect those testswhose results are invariant under a certain transformation. To observe the correlation, we use two methods. One is thedirect correlation between the tests and the other is the correlation between the results of a test on the sequence andits transformed form. In light of the observations, we conclude that some of the tests are correlated with each other.Furthermore, we conclude that in designing a reliable and efficient suite we can avoid overpopulating the list of testfunctions by employing transformations together with a reasonable number of statistical test functions.
