An affinity-based new local distance function and similarity measure for kNN algorithm

In this paper, we propose a modified version of the k-nearest neighbor (kNN) algorithm. We first introduce a new affinity function for distance measure between a test point and a training point which is an approach based on local learning. A new similarity function using this affinity function is proposed next for the classification of the test patterns. The widely used convention of k, i.e., k = [√N] is employed, where N is the number of data used for training purpose. The proposed modified kNN algorithm is applied on fifteen numerical datasets from the UCI machine learning data repository. Both 5-fold and 10-fold cross-validations are used. The average classification accuracy, obtained from our method is found to exceed some well-known clustering algorithms.     

An agglomerative clustering algorithm using a dynamic k-nearest-neighbor list

In this paper, a new algorithm is developed to reduce the computational complexity of Ward’s method. The proposed approach uses a dynamic k-nearest-neighbor list to avoid the determination of a cluster’s nearest neighbor at some steps of the cluster merge. Double linked algorithm (DLA) can significantly reduce the computing time of the fast pairwise nearest neighbor (FPNN) algorithm by obtaining an approximate solution of hierarchical agglomerative clustering. In this paper, we propose a method to resolve the problem of a non-optimal solution for DLA while keeping the corresponding advantage of low computational complexity. The computational complexity of the proposed method DKNNA + FS (dynamic k-nearest-neighbor algorithm with a fast search) in terms of the number of distance calculations is O(N²), where N is the number of data points. Compared to FPNN with a fast search (FPNN + FS), the proposed method using the same fast search algorithm (DKNNA + FS) can reduce the computing time by a factor of 1.90-2.18 for the data set from a real image. In comparison with FPNN + FS, DKNNA + FS can reduce the computing time by a factor of 1.92-2.02 using the data set generated from three images. Compared to DLA with a fast search (DLA + FS), DKNNA + FS can decrease the average mean square error by 1.26% for the same data set.     

A simplified multi-class support vector machine with reduced dual optimization

Support vector machine (SVM) was initially designed for binary classification. To extend SVM to the multi-class scenario, a number of classification models were proposed such as the one by Crammer and Singer (2001). However, the number of variables in Crammer and Singer’s dual problem is the product of the number of samples (l) by the number of classes (k), which produces a large computational complexity. This paper presents a simplified multi-class SVM (SimMSVM) that reduces the size of the resulting dual problem from l × k to l by introducing a relaxed classification error bound. The experimental results demonstrate that the proposed SimMSVM approach can greatly speed-up the training process, while maintaining a competitive classification accuracy.     

Efficient bit-parallel multipliers over finite fields GF(2)

Hardware implementation of multiplication in finite field GF(2^m) based on sparse polynomials is found to be advantageous in terms of space-complexity as well as the time-complexity. In this paper, we present a new permutation method to construct the irreducible like-trinomials of the form (x + 1)^m + (x + 1)ⁿ + 1 for the implementation of efficient bit-parallel multipliers. For implementing the multiplications based on such polynomials, we have defined a like-polynomial basis (LPB) as an alternative to the original polynomial basis of GF(2^m). We have shown further that the modular arithmetic for the binary field based on like-trinomials is equivalent to the arithmetic for the field based on trinomials. In order to design multipliers for composite fields, we have found another permutation polynomial to convert irreducible polynomials into like-trinomials of the forms (x² + x + 1)^m + (x² + x + 1)ⁿ + 1, (x² + x)^m + (x² + x)ⁿ + 1 and (x⁴ + x + 1)^m + (x⁴ + x + 1)ⁿ + 1. The proposed bit-parallel multiplier over GF(2^4m) is found to offer a saving of about 33% multiplications and 42.8% additions over the corresponding existing architectures.     

Estimation of instantaneous net surface longwave radiation from MODIS cloud-free data

This paper develops a statistical regression method to estimate the instantaneous Downwelling Surface Longwave Radiation (DSLR) for cloud-free skies using only the satellite-based radiances measured at the Top Of the Atmosphere (TOA), and subsequently combines the DSLR with the MODIS land surface temperature/emissivity products (MOD11_L2) to estimate the instantaneous Net Surface Longwave Radiation (NSLR). The proposed method relates the DSLR directly to the TOA radiances in the MODIS Thermal InfraRed (TIR) channels provided that the terrain altitude and the satellite Viewing Zenith Angle (VZA) are known. The simulation analysis shows that the instantaneous DSLR could be estimated by the proposed method with the Root Mean Square Error (RMSE) of 12.4 W/m² for VZA = 0 and terrain altitude z = 0 km. Similar results are obtained for the other VZAs and altitudes. Considering the MODIS instrumental errors of 0.25 K for the TOA brightness temperatures in channels 28, 33 and 34, and of 0.05 K for channels 29 and 31, and of 0.35 K for channel 36, the overall retrieval accuracy in terms of the RMSE is decreased to 13.1 W/m² for the instantaneous DSLR. Moreover, a comparison of MODIS derived DSLR and NSLR are done with the field measurements made at six sites of the Surface Radiation Budget Network (SURFRAD) in the United States for days with cloud-free conditions at the moment of MODIS overpass in 2006. The results show that the bias, RMSE and the square of the correlation coefficient (R²) between the MODIS derived DSLR with the proposed method and the field measured DSLR are 20.3 W/m², 30.1 W/m² and 0.91 respectively, and bias = 11.7 W/m², RMSE = 26.1 W/m² and R² = 0.94 for NSLR. In addition, the scheme proposed by Bisht et al. [Bisht, G., Venturini, V., Islam, S., & Jiang, L. (2005). Estimation of the net radiation using MODIS (Moderate Resolution Imaging Spectroradiometer) data for clear-sky days. Remote Sensing of Environment, 97, 52-67], which requires the MODIS atmospheric profile product (MOD07) and also the MODIS land surface temperature/emissivity products (MOD11_L2) as inputs, is used to estimate the instantaneous DSLR and NSLR for comparison with the field measurements as well as the MODIS derived DSLR and NSLR using our proposed method. The results of the comparisons show that, at least for our cases, our proposed method for estimating DSLR from the MODIS radiances at the TOA and the resultant NSLR gives results comparable to those estimated with Bisht et al.'s scheme [Bisht, G., Venturini, V., Islam, S., & Jiang, L. (2005). Estimation of the net radiation using MODIS (Moderate Resolution Imaging Spectroradiometer) data for clear-sky days. Remote Sensing of Environment, 97, 52-67].     

Adaptive reversible image watermarking scheme

This paper presents an adaptive block sized reversible image watermarking scheme. A reversible watermarking approach recovers the original image from a watermarked image after extracting the embedded watermarks. Without loss of generality, the proposed scheme segments an image of size 2^N × 2^N adaptively to blocks of size 2^L × 2^L, where L starts from a user-defined number to 1, according to their block structures. If possible, the differences between central ordered pixel and other pixels in each block are enlarged to embed watermarks. The embedded quantity is determined by the largest difference in a block and watermarks are embedded into LSB bits of above differences. Experimental results show that the proposed adaptive block size scheme has higher capacity than conventional fixed block sized method.     

Fast, prime factor, discrete Fourier transform algorithms over GF(2) for 8 ? m ? 10

In this paper it is shown that Winograd’s algorithm for computing convolutions and a fast, prime factor, discrete Fourier transform (DFT) algorithm can be modified to compute Fourier-like transforms of long sequences of 2^m − 1 points over GF(2^m), for 8 ? m ? 10. These new transform techniques can be used to decode Reed-Solomon (RS) codes of block length 2^m − 1. The complexity of this new transform algorithm is reduced substantially from more conventional methods. A computer simulation verifies these new results.     

Flow and thermal properties in full-zone thermocapillary liquid bridges

Properties of low Prandtl number flows are derived for slender cylindrical liquid bridges driven by interfacial thermocapillary forces. The theory and computation address the full-zone setting with both ‘outward’ (positive Marangoni number) and ‘inward’ (negative) flows along the liquid-gas interface. Three main results are found. The first concerns three-dimensional nonlinear mode interaction. Full numerical simulation reveals the unsteady three-dimensional nature of the flow solution beyond a cut-off value. Attention is paid to the case where the radius-to-height aspect ratio is 0.5, from which the nonlinearly-coupled azimuthal modes m = 1 and m = 2 are seen to dominate. Second is the behaviour for negative Marangoni number, where asymptotic analysis reveals that a critical value of the scaled Marangoni number exists, on approach to which the pressure gradient across the middle of the zone becomes large and negative. Computational solutions show encouraging agreement with asymptotic predictions (as well as slender-flow midzone computations) near that critical negative value. Third, both the negative and the positive critical Marangoni numbers are seen to correspond to the onset of ‘lemonheads’ (non-convex radial velocity profiles near the midzone) which directly promote three-dimensional instability.     

Remote estimation of chl-a concentration in turbid productive waters — Return to a simple two-band NIR-red model?

Today the water quality of many inland and coastal waters is compromised by cultural eutrophication in consequence of increased human agricultural and industrial activities. Remote sensing is widely applied to monitor the trophic state of these waters. This study investigates the performance of near infrared-red models for the remote estimation of chlorophyll-a concentrations in turbid productive waters and evaluates several near infrared-red models developed within the last 34 years. Three models were calibrated for a dataset with chlorophyll-a concentrations from 0 to 100 mg m⁻³ and validated for independent and statistically different datasets with chlorophyll-a concentrations from 0 to 100 mg m⁻³ and 0 to 25 mg m⁻³ for the spectral bands of the MEdium Resolution Imaging Spectrometer (MERIS) and MODerate resolution Imaging Spectroradiometer (MODIS). The MERIS two-band model estimated chlorophyll-a concentrations slightly more accurately than the more complex models, with mean absolute errors of 2.3 mg m⁻³ for chlorophyll-a concentrations from 0 to 100 mg m⁻³ and 1.2 mg m⁻³ for chlorophyll-a concentrations from 0 to 25 mg m⁻³. Comparable results from several near infrared-red models with different levels of complexity, calibrated for inland and coastal waters around the world, indicate a high potential for the development of a simple universally applicable near infrared-red algorithm.     

Hashing via finite field

In order to produce full length probe sequences, the table size m for many existing open addressing hash functions, for example, the widely used double hashing, must be prime, i.e., m = p where p is prime. In this paper, we propose a new and efficient open addressing technique, called hashing via finite field, to construct a new class of hash functions with table size m = pⁿ, where p is prime and n ? 1. It is clear that it includes prime m as a special case when n = 1. We show that the new class of hash functions constructed via finite field produces full length probe sequences on all table elements. Also, some theoretic analysis is provided along with concrete examples.     

Machine learning approach to model sport training

The aim of this study was to use a machine learning approach combining fuzzy modeling with an immune algorithm to model sport training, in particular swimming. A proposed algorithm mines the available data and delivers the results in a form of a set of fuzzy rules “IF (fuzzy conditions) THEN (class)”. Fuzzy logic is a powerful method to cope with continuous data, to overcome problem of overlapping class definitions, and to improve the rule comprehensibility. Sport training is modeled at the level of microcycle and training unit by 12 independent attributes. The data was collected in two months (February-March 2008), among swimmers from swimming sections in Wroc?aw, Poland. The swimmers had minimum of 7 years of training and reached the II class level in swimming classification from 2005 to 2008. The goal of the performed experiments was to find the rules answering the question - how does the training unit influence swimmer’s feelings while being in water the next day? The fuzzy rules were inferred for two different scales of the class to be predicted. The effectiveness of the learned set of rules reached 68.66%. The performance, in terms of classification accuracy, of the proposed approach was compared with traditional classifier schemes. The accuracy of the result of compared methods is significantly lower than the accuracy of fuzzy rules obtained by a method presented in this study (paired t-test, P < 0.05).     

An assessment of chlorophyll-a algorithms available for SeaWiFS in coastal and open areas of the Bay of Bengal and Arabian Sea

Three ocean colour algorithms, OC4v6, Carder and OC5 were tested for retrieving Chlorophyll-a (Chla) in coastal areas of the Bay of Bengal and open ocean areas of the Arabian Sea. Firstly, the algorithms were run using ~ 80 in situ Remote Sensing Reflectance, (R_rs(λ)) data collected from coastal areas during eight cruises from January 2000 to March 2002 and the output was compared to in situ Chla. Secondly, the algorithms were run with ~ 20 SeaWiFS R_rs(λ) and the results were compared with coincident in situ Chla. In both cases, OC5 exhibited the lowest log₁₀-RMS, bias, had a slope close to 1 and this algorithm appears to be the most accurate for both coastal and open ocean areas. Thirdly the error in the algorithms was regressed against Total Suspended Material (TSM) and Coloured Dissolved Organic Material (CDOM) data to assess the co-variance with these parameters. The OC5 error did not co-vary with TSM and CDOM. OC4v6 tended to over-estimate Chla > 2 mg m⁻³ and the error in OC4v6 co-varied with TSM. OC4v6 was more accurate than the Carder algorithm, which over-estimated Chla at concentrations > 1 mg m⁻³ and under-estimated Chla at values < 0.5 mg m⁻³. The error in Carder Chla also co-varied with TSM. The algorithms were inter-compared using > 5500 SeaWiFS R_rs(λ) data from coastal to offshore transects in the Northern Bay of Bengal. There was good agreement between OC4v6 and OC5 in open ocean waters and in coastal areas up to 2 mg m⁻³. There was a strong divergence between Carder and OC5 in open ocean and coastal waters. OC4v6 and Carder tended to over-estimate Chla in coastal areas by a factor of 2 to 3 when TSM > 25 g m⁻³. We strongly recommend the use of OC5 for coastal and open ocean waters of the Bay of Bengal and Arabian Sea. A Chla time series was generated using OC5 from 2000 to 2003, which showed that concentrations at the mouths of the Ganges reach a maxima (~ 5 mg m⁻³) in October and November and were 0.08 mg m⁻³ further offshore increasing to 0.2 mg m⁻³ during December. Similarly in early spring from February to March, Chla was 0.08 to 0.2 mg m⁻³ on the east coast of the Bay.     

A discrete geometry approach for dominant point detection

We propose two fast methods for dominant point detection and polygonal representation of noisy and possibly disconnected curves based on a study of the decomposition of the curve into the sequence of maximal blurred segments [2]. Starting from results of discrete geometry [3] and [4], the notion of maximal blurred segment of width ν[2] has been proposed, well adapted to possibly noisy curves. The first method uses a fixed parameter that is the width of considered maximal blurred segments. The second method is deduced from the first one based on a multi-width approach to obtain a non-parametric method that uses no threshold for working with noisy curves. Comparisons with other methods in the literature prove the efficiency of our approach. Thanks to a recent result [5] concerning the construction of the sequence of maximal blurred segments, the complexity of the proposed methods is O(n log n). An application of vectorization is also given in this paper.     

A new technique for extracting the red edge position from hyperspectral data: The linear extrapolation method

The position of the inflexion point in the red edge region (680 to 780 nm) of the spectral reflectance signature, termed the red edge position (REP), is affected by biochemical and biophysical parameters and has been used as a means to estimate foliar chlorophyll or nitrogen content. In this paper, we report on a new technique for extracting the REP from hyperspectral data that aims to mitigate the discontinuity in the relationship between the REP and the nitrogen content caused by the existence of a double-peak feature on the derivative spectrum. It is based on a linear extrapolation of straight lines on the far-red (680 to 700 nm) and NIR (725 to 760 nm) flanks of the first derivative reflectance spectrum. The REP is then defined by the wavelength value at the intersection of the two lines. The output is a REP equation, REP = − (c₁ − c₂) / (m₁ − m₂), where c₁ and c₂, and m₁ and m₂ represent the intercepts and slopes of the far-red and NIR lines, respectively. Far-red wavebands at 679.65 and 694.30 nm in combination with NIR wavebands at 732.46 and 760.41 nm or at 723.64 and 760.41 nm were identified as the optimal combinations for calculating nitrogen-sensitive REPs for three spectral data sets (rye canopy, and maize leaf and mixed grass/herb leaf stack spectra). REPs extracted using this new technique (linear extrapolation method) showed high correlations with a wide range of foliar nitrogen concentrations for both narrow and wider bandwidth spectra, being comparable with results obtained using the traditional linear interpolation, polynomial and inverted Gaussian fitting techniques. In addition, the new technique is simple as is the case with the linear interpolation method, but performed better than the latter method in the case of maize leaves at different developmental stages and mixed grass/herb leaves with a low nitrogen concentration.     

Red tide detection and tracing using MODIS fluorescence data: A regional example in SW Florida coastal waters

Near real-time data from the MODIS satellite sensor was used to detect and trace a harmful algal bloom (HAB), or red tide, in SW Florida coastal waters from October to December 2004. MODIS fluorescence line height (FLH in W m^− 2 μm^− 1 sr^− 1) data showed the highest correlation with near-concurrent in situ chlorophyll-a concentration (Chl in mg m^− 3). For Chl ranging between 0.4 to 4 mg m^− 3 the ratio between MODIS FLH and in situ Chl is about 0.1 W m^− 2 μm^− 1 sr^− 1 per mg m^− 3 chlorophyll (Chl = 1.255 (FLH × 10)^0.86, r = 0.92, n = 77). In contrast, the band-ratio chlorophyll product of either MODIS or SeaWiFS in this complex coastal environment provided false information. Errors in the satellite Chl data can be both negative and positive (3-15 times higher than in situ Chl) and these data are often inconsistent either spatially or temporally, due to interferences of other water constituents. The red tide that formed from November to December 2004 off SW Florida was revealed by MODIS FLH imagery, and was confirmed by field sampling to contain medium (10⁴ to 10⁵ cells L^− 1) to high (> 10⁵ cells L^− 1) concentrations of the toxic dinoflagellate Karenia brevis. The FLH imagery also showed that the bloom started in mid-October south of Charlotte Harbor, and that it developed and moved to the south and southwest in the subsequent weeks. Despite some artifacts in the data and uncertainty caused by factors such as unknown fluorescence efficiency, our results show that the MODIS FLH data provide an unprecedented tool for research and managers to study and monitor algal blooms in coastal environments.     

Reliability of LiDAR derived predictors of forest inventory attributes: A case study with Norway spruce

To increase the application domain (re-use) of LiDAR-based models the random replication effects in the predictor(s) must be considered. We quantify these effects in a linear predictor (X) of four forest inventory attributes (Lorey's height HT, basal area BA, volume VOL, and stem density TPH) with LiDAR data acquired over 40 spruce-dominated large plots in southeastern Norway. A grid-based random thinning of the raw multi-echo LiDAR data, to five target densities between 0.25 m^− 2 and 2.0 m^− 2, generated 100 replications with each density. A DTM was estimated for each replicate and target pulse density. The four linear predictors were constructed from two indicators of canopy density and a posited average effect of a power-transform of echoes classified as canopy returns. Replication variance varied significantly among plots but the reliability ratio of X was high (≥ 0.92) for HT, BA and VOL but lower for TPH, especially at low pulse densities. Reliability ratios increased with pulse density. Replication variance attenuated the linear regression coefficients by about 10% and inflated the residual variance by 3-6%. A proposed calibration was effective in reducing the impact of replication effects. A proposed bootstrap procedure can be used in practice to obtain good approximations of the replication variance. With echo-densities of approximately 1 m^− 2 or higher the replication effects do not warrant the effort of a calibration.     

Ship motions using single-phase level set with dynamic overset grids

The problem of surface ships free to pitch and heave in regular head waves is analyzed numerically with an unsteady Reynolds averaged Navier Stokes (URANS) approach. The unsteady single-phase level set method previously developed by the authors was extended to include six degrees of freedom (6DOF) motions. The method uses rigid overset grids that move with relative motion during the computation, and the interpolation coefficients between the grids are recomputed dynamically every time the grids move. The motions in each time step are integrated implicitly using a predictor-corrector approach. An earth-based reference system is used for the solution of the fluid flow, while a ship-based reference system is used to compute the rigid-body equations of motion. Predicted results for sinkage and trim and resistance at two Froude numbers (medium, Fr = 0.28 and large, Fr = 0.41) were compared against experimental data, showing good agreement. Pitch and heave motions were computed for near-resonant cases at Fr = 0.28 and 0.41, with regular linear head waves with slope ak = 0.025 and wavelength λ = 1.5L, with L the ship length. The predicted motions compare favorably with existing experimental data. A solution for a large amplitude head wave case (ak = 0.075) was also obtained, in which the transom wave breaks and extreme motions are observed. The medium Froude number case was subject to a verification and validation analysis. A problem with two ships pitching and heaving one behind the other is demonstrated.     

Regional evaporation estimates from flux tower and MODIS satellite data

Two models were evaluated for their ability to estimate land surface evaporation at 16-day intervals using MODIS remote sensing data and surface meteorology as inputs. The first was the aerodynamic resistance-surface energy balance model, and the second was the Penman-Monteith (P-M) equation, where the required surface conductance is estimated from remotely-sensed leaf area index. The models were tested using 3 years of evaporation and meteorological measurements from two contrasting Australian ecosystems, a cool temperate, evergreen Eucalyptus forest and a wet/dry, tropical savanna. The aerodynamic resistance-surface energy balance approach failed because small errors in the radiative surface temperature translate into large errors in sensible heat, and hence into estimates of evaporation. The P-M model adequately estimated the magnitude and seasonal variation in evaporation in both ecosystems (RMSE = 27 W m^− 2, R² = 0.74), demonstrating the validity of the proposed surface conductance algorithm. This, and the ability to constrain evaporation estimates via the energy balance, demonstrates the superiority of the P-M equation over the surface temperature-based model. There was no degradation in the performance of the P-M model when gridded meteorological data at coarser spatial (0.05°) and temporal (daily) resolution were substituted for locally-measured inputs.The P-M approach was used to generate a monthly evaporation climatology for Australia from 2001 to 2004 to demonstrate the potential of this approach for monitoring land surface evaporation and constructing monthly water budgets from 1-km to continental spatial scales.     

Recursive algorithm based on fuzzy 2-partition entropy for 2-level image thresholding

The fuzzy c-partition entropy approach for threshold selection behaves well in segmenting images. But the size of search space increases very rapidly when the number of parameters needed to determine the membership function increases. The computation complexity of the fuzzy 2-partition entropy approach is bounded by O(L³). In this paper, a recursive scheme which decreases the computation complexity of the basic algorithm to O(L²) is proposed. The approach does not need the calculation of the membership function. The processing time of each image is reduced from more than 5 min to less than 20 s.