scholarly journals Time series classification using k-Nearest neighbours, Multilayer Perceptron and Learning Vector Quantization algorithms

2012 ◽  
Vol 60 (2) ◽  
pp. 69-72 ◽  
Author(s):  
Jiří Fejfar ◽  
Jiří Šťastný ◽  
Miroslav Cepl
Keyword(s):  
Time Series ◽  
Vector Quantization ◽  
Multilayer Perceptron ◽  
Confusion Matrix ◽  
A Priori ◽  
Learning Vector Quantization ◽  
Self Organizing Map ◽  
Data Set ◽  
Nearest Neighbours ◽  
Musical Excerpts

We are presenting results comparison of three artificial intelligence algorithms in a classification of time series derived from musical excerpts in this paper. Algorithms were chosen to represent different principles of classification – statistic approach, neural networks and competitive learning. The first algorithm is a classical k-Nearest neighbours algorithm, the second algorithm is Multilayer Perceptron (MPL), an example of artificial neural network and the third one is a Learning Vector Quantization (LVQ) algorithm representing supervised counterpart to unsupervised Self Organizing Map (SOM).After our own former experiments with unlabelled data we moved forward to the data labels utilization, which generally led to a better accuracy of classification results. As we need huge data set of labelled time series (a priori knowledge of correct class which each time series instance belongs to), we used, with a good experience in former studies, musical excerpts as a source of real-world time series. We are using standard deviation of the sound signal as a descriptor of a musical excerpts volume level.We are describing principle of each algorithm as well as its implementation briefly, giving links for further research. Classification results of each algorithm are presented in a confusion matrix showing numbers of misclassifications and allowing to evaluate overall accuracy of the algorithm. Results are compared and particular misclassifications are discussed for each algorithm. Finally the best solution is chosen and further research goals are given.

scholarly journals Multipose Face Recognition-Based Combined Adaptive Deep Learning Vector Quantization

2020 ◽  
Vol 2020 ◽  
pp. 1-11
Author(s):  
Shahenda Sarhan ◽  
Aida A. Nasr ◽  
Mahmoud Y. Shams
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Face Recognition ◽  
Vector Quantization ◽  
Confusion Matrix ◽  
Recognition System ◽  
Learning Vector Quantization ◽  
Majority Voting ◽  
Support Vector ◽  
Reliability And Robustness

Multipose face recognition system is one of the recent challenges faced by the researchers interested in security applications. Different researches have been introduced discussing the accuracy improvement of multipose face recognition through enhancing the face detector as Viola-Jones, Real Adaboost, and Cascade Object Detector while others concentrated on the recognition systems as support vector machine and deep convolution neural networks. In this paper, a combined adaptive deep learning vector quantization (CADLVQ) classifier is proposed. The proposed classifier has boosted the weakness of the adaptive deep learning vector quantization classifiers through using the majority voting algorithm with the speeded up robust feature extractor. Experimental results indicate that, the proposed classifier provided promising results in terms of sensitivity, specificity, precision, and accuracy compared to recent approaches in deep learning, statistical, and classical neural networks. Finally, the comparison is empirically performed using confusion matrix to ensure the reliability and robustness of the proposed system compared to the state-of art.

scholarly journals Perbandingan Algoritma LVQ dan RBFNN Untuk Identifikasi Glaukoma dan Diabetes Retinopati Pada Citra Fundus

2020 ◽  
Vol 1 (1) ◽  
pp. 68-77
Author(s):  
Kevin Oktavius ◽  
Siska Devella
Keyword(s):  
Neural Network ◽  
Radial Basis Function ◽  
Vector Quantization ◽  
Basis Function ◽  
Confusion Matrix ◽  
Learning Vector Quantization ◽  
Radial Basis

Penyakit mata merupakan salah satu masalah kesehatan utama pada semua orang terutama pada kaum lansia, penyakit mata yang paling umum menyerang lansia diantaranya adalah glaukoma dan retinopati diabetes. Penyakit glaukoma dan diabetes retinopati dapat diketahui melalui citra fundus. Pada penelitian ini telah dilakukan perbandingan algoritma Learning Vector Quantization dengan Radial Basis Function Neural Network untuk klasifikasi penyakit glaukoma dan diabetes retinopati (accuracy, precision, recall) berdasarkan citra fundus resolusi tinggi. Dataset yang digunakan berjumlah 45 citra fundus yang terdiri dari 15 citra fundus terjangkit glaukoma, 15 citra fundus terjangkit diabetes retinopati dan 15 citra fundus mata normal. Pada perhitungan dengan confusion matrix hasil tertinggi didapatkan pada algoritma radial basis function neural network dengan spread=20 dan MN=10 menghasilkan rata-rata accuracy sebesar 81,06%, precision sebesar 80,83% dan recall sebesar 73,33% jika dibandingkan dengan algoritma learning vector quantization dengan lvqnet=50 dan epoch=45 menghasilkan rata-rata accuracy sebesar 80,85%, precision sebesar 73,33% dan recall sebesar 77,14%.

scholarly journals Five years of CO, HCN, C<sub>2</sub>H<sub>6</sub>, C<sub>2</sub>H<sub>2</sub>, CH<sub>3</sub>OH, HCOOH and H<sub>2</sub>CO total columns measured in the Canadian high Arctic

2014 ◽  
Vol 7 (6) ◽  
pp. 1547-1570 ◽  
Author(s):  
C. Viatte ◽  
K. Strong ◽  
K. A. Walker ◽  
J. R. Drummond
Keyword(s):  
Time Series ◽  
Fourier Transform ◽  
Biomass Burning ◽  
Atmospheric Chemistry ◽  
A Priori ◽  
Estimation Method ◽  
Optimal Estimation ◽  
High Arctic ◽  
Retrieval Algorithm ◽  
Data Set

Abstract. We present a five-year time series of seven tropospheric species measured using a ground-based Fourier transform infrared (FTIR) spectrometer at the Polar Environment Atmospheric Research Laboratory (PEARL; Eureka, Nunavut, Canada; 80°05' N, 86°42' W) from 2007 to 2011. Total columns and temporal variabilities of carbon monoxide (CO), hydrogen cyanide (HCN) and ethane (C2H6) as well as the first derived total columns at Eureka of acetylene (C2H2), methanol (CH3OH), formic acid (HCOOH) and formaldehyde (H2CO) are investigated, providing a new data set in the sparsely sampled high latitudes. Total columns are obtained using the SFIT2 retrieval algorithm based on the optimal estimation method. The microwindows as well as the a priori profiles and variabilities are selected to optimize the information content of the retrievals, and error analyses are performed for all seven species. Our retrievals show good sensitivities in the troposphere. The seasonal amplitudes of the time series, ranging from 34 to 104%, are captured while using a single a priori profile for each species. The time series of the CO, C2H6 and C2H2 total columns at PEARL exhibit strong seasonal cycles with maxima in winter and minima in summer, in opposite phase to the HCN, CH3OH, HCOOH and H2CO time series. These cycles result from the relative contributions of the photochemistry, oxidation and transport as well as biogenic and biomass burning emissions. Comparisons of the FTIR partial columns with coincident satellite measurements by the Atmospheric Chemistry Experiment Fourier Transform Spectrometer (ACE-FTS) show good agreement. The correlation coefficients and the slopes range from 0.56 to 0.97 and 0.50 to 3.35, respectively, for the seven target species. Our new data set is compared to previous measurements found in the literature to assess atmospheric budgets of these tropospheric species in the high Arctic. The CO and C2H6concentrations are consistent with negative trends observed over the Northern Hemisphere, attributed to fossil fuel emission decrease. The importance of poleward transport for the atmospheric budgets of HCN and C2H2 is highlighted. Columns and variabilities of CH3OH and HCOOH at PEARL are comparable to previous measurements performed at other remote sites. However, the small columns of H2CO in early May might reflect its large atmospheric variability and/or the effect of the updated spectroscopic parameters used in our retrievals. Overall, emissions from biomass burning contribute to the day-to-day variabilities of the seven tropospheric species observed at Eureka.

scholarly journals Maximum Fraction Images Derived from Year-Based Project for On-Board Autonomy-Vegetation (PROBA-V) Data for the Rapid Assessment of Land Use and Land Cover Areas in Mato Grosso State, Brazil

Land ◽  
2020 ◽  
Vol 9 (5) ◽  
pp. 139 ◽  
Author(s):  
Henrique Luis Godinho Cassol ◽  
Egidio Arai ◽  
Edson Eyji Sano ◽  
Andeise Cerqueira Dutra ◽  
Tânia Beatriz Hoffmann ◽  
...  
Keyword(s):  
Land Use ◽  
Time Series ◽  
Land Cover ◽  
Confusion Matrix ◽  
Google Earth ◽  
Landsat 8 ◽  
Validation Data ◽  
Data Set ◽  
Mato Grosso ◽  
Fraction Images

This paper presents a new approach for rapidly assessing the extent of land use and land cover (LULC) areas in Mato Grosso state, Brazil. The novel idea is the use of an annual time series of fraction images derived from the linear spectral mixing model (LSMM) instead of original bands. The LSMM was applied to the Project for On-Board Autonomy-Vegetation (PROBA-V) 100-m data composites from 2015 (~73 scenes/year, cloud-free images, in theory), generating vegetation, soil, and shade fraction images. These fraction images highlight the LULC components inside the pixels. The other new idea is to reduce these time series to only six single bands representing the maximum and standard deviation values of these fraction images in an annual composite, reducing the volume of data to classify the main LULC classes. The whole image classification process was conducted in the Google Earth Engine platform using the pixel-based random forest algorithm. A set of 622 samples of each LULC class was collected by visual inspection of PROBA-V and Landsat-8 Operational Land Imager (OLI) images and divided into training and validation datasets. The performance of the method was evaluated by the overall accuracy and confusion matrix. The overall accuracy was 92.4%, with the lowest misclassification found for cropland and forestland (<9% error). The same validation data set showed 88% agreement with the LULC map made available by the Landsat-based MapBiomas project. This proposed method has the potential to be used operationally to accurately map the main LULC areas and to rapidly use the PROBA-V dataset at regional or national levels.

scholarly journals Selection of number of neurons for vector quantization methods

2008 ◽  
Vol 48 ◽  
Author(s):  
Olga Kurasova ◽  
Alma Molytė
Keyword(s):  
Neural Networks ◽  
Vector Quantization ◽  
Self Organizing Map ◽  
Data Set ◽  
Selection Of ◽  
Self Organizing

In this paper, a strategy of the selection of the neurons number for vector quantization methods has been investigated. Two methods based on neural networks have been analysed: self-organizing map and neuralgas. There is suggested a way under which the number of neurons is selected taken into account the particularity of the analysed data set.

scholarly journals PENGENALAN POLA SIDIK JARI DENGAN METODE LOCAL BINARY PATTERN DAN LEARNING VECTOR QUANTIZATION

2019 ◽  
Vol 7 (2) ◽  
pp. 148-156
Author(s):  
Adriana Fanggidae ◽  
Dony M Sihotang ◽  
Adnan Putra Rihi Pati
Keyword(s):  
Vector Quantization ◽  
Local Binary Pattern ◽  
Learning Vector Quantization ◽  
Data Set ◽  
Otsu Thresholding

Sidik jari merupakan strukur genetika dalam bentuk pola yang sangat detail dan tanda yang melekat pada diri manusia. Banyak sistem biometrika yang menggunakan sidik jari sebagai data masukan, karena sifat dari sidik jari setiap individu berbeda meskipun kembar identik dan tidak berubah kecuali mendapat kecelakaan. Metode yang digunakan dalam penelitian ini yaitu segmentasi dengan algoritma Otsu thresholding, ekstraksi ciri dengan algoritma Local Binary Pattern (LBP), dan pembelajaran dengan algoritma Learning Vector Quantization (LVQ). Data yang digunakan adalah citra sidik jari jempol berukuran 200 x 300 piksel, berjenis keabuan dan berformat *.jpg. Citra sidik jari terdiri dari 25 orang, masing-masing orang memiliki 6 data latih dan 2 data uji. Pengujian data latih dan data uji dilakukan kepada empat sistem yaitu sistem dengan jumlah ciri LBP = 8, 64, 128 dan 256 dan menggunakan masing-masing 2 buah data set dimana data set 1 berjumlah 15 orang dan data set 2 berjumlah 25 orang. Hasil pengujian keempat sistem menunjukkan bahwa sistem dengan jumlah ciri LBP = 128 merupakan sistem yang terbaik dengan kombinasi akurasi sistem yang tinggi dan juga waktu pembelajaran yang cepat.

scholarly journals First intercalibration of column-averaged methane from the Total Carbon Column Observing Network and the Network for the Detection of Atmospheric Composition Change

2013 ◽  
Vol 6 (2) ◽  
pp. 397-418 ◽  
Author(s):  
R. Sussmann ◽  
A. Ostler ◽  
F. Forster ◽  
M. Rettinger ◽  
N. M. Deutscher ◽  
...  
Keyword(s):  
Time Series ◽  
Near Infrared ◽  
A Priori ◽  
Time Dependent ◽  
Total Carbon ◽  
Atmospheric Composition ◽  
Composition Change ◽  
Data Set ◽  
The Difference ◽  
Difference Time

Abstract. We present the first intercalibration of dry-air column-averaged mole fractions of methane (XCH4) retrieved from solar Fourier transform infrared (FTIR) measurements of the Network for the Detection of Atmospheric Composition Change (NDACC) in the mid-infrared (MIR) versus near-infrared (NIR) soundings from the Total Carbon Column Observing Network (TCCON). The study uses multi-annual quasi-coincident MIR and NIR measurements from the stations Garmisch, Germany (47.48° N, 11.06° E, 743 m a.s.l.), and Wollongong, Australia (34.41° S, 150.88° E, 30 m a.s.l.). Direct comparison of the retrieved MIR and NIR XCH4 time series for Garmisch shows a quasi-periodic seasonal bias leading to a standard deviation (stdv) of the difference time series (NIR–MIR) of 7.2 ppb. After reducing time-dependent a priori impact by using realistic site- and time-dependent ACTM-simulated profiles as a common prior, the seasonal bias is reduced (stdv = 5.2 ppb). A linear fit to the MIR/NIR scatter plot of monthly means based on same-day coincidences does not show a y-intercept that is statistically different from zero, and the MIR/NIR intercalibration factor is found to be close to ideal within 2-σ uncertainty, i.e. 0.9996(8). The difference time series (NIR–MIR) do not show a significant trend. The same basic findings hold for Wollongong. In particular an overall MIR/NIR intercalibration factor close to the ideal 1 is found within 2-σ uncertainty. At Wollongong the seasonal cycle of methane is less pronounced and corresponding smoothing errors are not as significant, enabling standard MIR and NIR retrievals to be used directly, without correction to a common a priori. Our results suggest that it is possible to set up a harmonized NDACC and TCCON XCH4 data set which can be exploited for joint trend studies, satellite validation, or the inverse modeling of sources and sinks.

scholarly journals Towards model evaluation and identification using Self-Organizing Maps

2008 ◽  
Vol 12 (2) ◽  
pp. 657-667 ◽  
Author(s):  
M. Herbst ◽  
M. C. Casper
Keyword(s):  
Time Series ◽  
Monte Carlo ◽  
Performance Measures ◽  
Model Identification ◽  
Equivalent Model ◽  
Self Organizing Map ◽  
Watershed Model ◽  
Self Organizing Maps ◽  
Data Set ◽  
Self Organizing

Abstract. The reduction of information contained in model time series through the use of aggregating statistical performance measures is very high compared to the amount of information that one would like to draw from it for model identification and calibration purposes. It has been readily shown that this loss imposes important limitations on model identification and -diagnostics and thus constitutes an element of the overall model uncertainty. In this contribution we present an approach using a Self-Organizing Map (SOM) to circumvent the identifiability problem induced by the low discriminatory power of aggregating performance measures. Instead, a Self-Organizing Map is used to differentiate the spectrum of model realizations, obtained from Monte-Carlo simulations with a distributed conceptual watershed model, based on the recognition of different patterns in time series. Further, the SOM is used instead of a classical optimization algorithm to identify those model realizations among the Monte-Carlo simulation results that most closely approximate the pattern of the measured discharge time series. The results are analyzed and compared with the manually calibrated model as well as with the results of the Shuffled Complex Evolution algorithm (SCE-UA). In our study the latter slightly outperformed the SOM results. The SOM method, however, yields a set of equivalent model parameterizations and therefore also allows for confining the parameter space to a region that closely represents a measured data set. This particular feature renders the SOM potentially useful for future model identification applications.

Sign in / Sign up

Export Citation Format

Share Document