Dew Lab Studio contains: MtxVec library with Stats Master, DSP Master and Data Miner add ons.
MtxVec feature list:
- Object oriented numerical library for Delphi and C++ Builder users and .NET with GPU Open CL and 64bit Delphi compiler support.
- Supports +,/,-,* for Vectors and Matrices executing faster than native compiler code.
- Code optimizations and support for Intel Core/Nehalem/Haswell product familiy.
- Simple functions like Sin, Cos.. are threaded when vector length exceeds about 5000 elements. No MtxVec code change required to have your code run on multiple cores.
- Complete support for complex numbers for all functions in single or double precision.
- Matrix operations: Solving a system of linear equations with refinment and condition number via LU, QR or SVD decomposition, Inverse of the matrix, Eigenvalues and eigenvectors, Sylvester equation, Teoplitz system solver, 2D FFT, Inverse 2D FFT, 2D real threaded FFT's, Square root of the matrix, Logarithm of the matrix...
- Vector operations: autocorrelation, circular convolution, cross correlation, forward and inverse discrete cosine transform, forward and inverse FFT, Geortzx92el algorithm, Hilbert transform...
- Probability distributions PDF, CDF and Inverse CDF: binomial, geometric, hypergeometric, negative binomial, Poisson, discrete uniform, beta ,Cauchy, chi-squared, exponential, F, gamma, Maxwell, normal, Pareto, Rayleigh, Student (T), Continuous uniform, Weibull..
- Sparse matrix operations: Solver for banded sparse matrices, Iterative sparse solvers: SGS, CG, BC, CS, GMRES, OM with LU preconditioning. Direct sparse solver for non-symmetric matrices. Conversions supported between: Banded, Dense, Sparse and Triplets matrix formats.
- Performance features: CPU specific optimizations, Critical functions optimized in assembler, symmetric multiprocessing, block processing support for vectors, fast object create and destroy, thin object layer, vector/matrix design, SSE4/AVX 2.0 enabled.
- Special functions: complete elliptic integrals, Jacoby elliptic functions, Airy function and its derivative, Airy function of second kind and its derivative, Bessel functions: J, Y, I, K and H, associated Legendre polynomials.
- Vectorized math parser capable of parsing equations with real and complex numbers, vectors and matrices. Customizable operators, functions and variables. Expressions evaluated on multiple cores.
- Polynomials: Finding roots, polynomial evaluation, Cubic splines, Linear and cubic interpolation algorithms, construction and evaluation of piece-wise polynoms, polynomial division (deconvolution) and multiplication (filter).
- Optimization methods: Brent, BFGS, Marquardt and Nelder-Mead minimization methods, numerical gradient and Hessian matrix and more...
- Numerical integration: Gauss quadrature, Trapezoid, Bool's, Simpson methods.
- Management functions: save and load matrices in different formats: binary, text, matrix-market. View matrices with built-in viewer. Visualize vectors and matrices on the charts, Store data in different precisions. Edit and display complex numbers, A thread component allows you to quickly provide efficient progress indicators and cancel option for your numerical algorithm.
- Sparse matrices: Umfpack v5.1 and updated Pardiso sparse solvers included.
- New HTML2 help file format.
- Compact MtxVec allows reduced distribution size with fewer dll's.
- debugger visualizer allows inspection and charting of vectors and matrices while debugging Open CL code.
- Ability to use single and double precision library concurrently (registered users only)
- Generalized eigenvalues including Schure Vectors, compute only few eigenvalues by range or count, symmetric or general, real or complex, for sparse or dense matrices.
- Generalized singular value decomposition.
- Solver for general Gauss-Markov linear model and for linear equality-constrained least squares
- Code runs on AMD, Nvidia and CPU devices
- Features 500 unique kernels and 2000 variants.
- Full real and complex vector math for single and double precision.
- Debugger visualizer support for vectors and matrices in GPU memory.
- Supports +,/,-,* for Vectors and Matrices located in GPU memory.
- Features separate kernels for GPU and CPU devices for maximum performance.
- Automatically precompiles code on first run and caches binaries for the future.
- Comprehensive object wrappers for most of the Open CL API.
- Dynamic Open CL dll library linking.
- CPU code fallback built in to the library.
- Tested with 32bit and 64bit compiler.
- Ready to use multi-threaded job distribution between GPU devices.
- Four AMD 6950 series devices running Cougar Open CL run specific algorithms in double precision complex math more than 3x faster than 8 way 10 core per CPU (80 core) Intel Xeon based server
- 19 different distributions (PDF, CDF and inverse CDF function), mean and variance for all 19 distributions, random generators for 15 distributions, parameter estimate for beta, binomial, exponential, gamma, geometric, normal, Poisson, continuous uniform and Weibull distribution.
- Descriptive statistics: histograms, ogives nth-Moment, percentile, range, Interquertile Range IQR, mean, harmonic mean, geometric mean, trimmed mean, median, skewness, kurtosis, standard deviation, root mean square RMS, minimum, maximum, covariance, correlation matrix, cumulative sum.
- Principal component analysis (PCA): PCA by using the covariance matrix, PCA Residuals, Barlett test for dimensionality, Hypothesys testing: Nonparametric tests (sign test, Wilcoxon tests)
- Parametric tests (one-sample t-test, two-sample paired/unpaired t-test), Z test, Regression models: linear (weighted, unweighted), multiple linear (weighted, unweighted), ridge regression, non-linear regression (using the BFGS, Marquardt or Simplex method) , one-way and two-way ANOVA.
- Optional install: Statistical charts (precondition - installed TeeChart v5 PRO) normal plot, QQ plot, XBar plot, SPS charts, Weibull plot, Box plot, Histogram plot. The package is built on top of MtxVec numerical library, and takes advantage of CPU specific optimizations, dynamic memory allocation and object oriented design.
- IIR filters: Design of analog or discrete Butterworth, ChebyshevI, ChebyshevII, Elliptic and Bessel filters. Order estimation includes all filter types: lowpass, highpass, bandpass and bandstop. Bilinear and Matched Z transform, Group delay, Frequency transformations in S or Z domain, in zero-pole or state-space form.
- FIR Filters: Design of FIR filters with window method or with remez exchange algorithm. Order estimation of FIR filters. Design of hilbert transformers, differentiators and also integrators. Savitzky-Golay filtering, fast envelope detector. Multi-rate multi-stage half-band FIR filtering support includes: zoom-spectrum component, decimate and interpolate components and a high quality envelope detector.
- Real factor rate converters, modulators, demodulators and envelope detectors.
- Non-linear filters: sample-and-hold, sample-and-decay and median filter.
- Spectral analysis: State-of-the-art spectrum analyzer component with ready to use component editor covering: FFT, CZT, Yule-Walker, Burg, Covariance and Modified Covariance spectral estimation, RMS of specified frequency bands, a set of peak interpolation algorithms, a large set of window functions including: Bartlett, Blackman, Chebyshev, CosineTappered, FlatTop, Hamming, Hanning and Kaiser window; sophisticated peak selection and peak tracking methods, phase unwrapping algorithm, dedicated components for bispectrum, bicoherence, coherence and transfer function estimation with real-time capability. Real cepstrum, complex cepstrum and inverse complex cepstrum, Spectral statistics: Noise floor, SFDR, THD, THDN, SINAD, RMS, SNR...
- Linear systems: find zeros of a linear system, convert between state-space, zero-pole and numerator-denominator forms of the transfer function.
- Other components: signal generator component with ready to use component editor offers stack-based vectorized function evaluation with many built-in functions. Audio recording and playback components with monitor function and extensive support for triggers. Ready to use component editor for filtering component covers design of nearly all included filters. Components for streaming (read/write) different file formats.
- TeeChart Pro support includes: two new TChart components to allow visual connection of the signal processing pipes to charts, two new TChartSeries for faster drawing and 3 additional TeeTools components to help with peak marking and axis scaling.
- Together with MtxVec the following functions/operations are provided: frequency response of analog and discrete filters, fast 1D and 2D convolution, fast autocorrelation and cross-correlation, deconvolution, 1D and 2D filtering, zero phase IIR filtering, interpolated FIR filters, DCT and inverse DCT, Geortzel algorithm and forward and inverse FFT for all combinations of real/complex source and destination, random number generators and more...
- Data miner is a set of components for classification for Delphi written in 100% VCL. The algorithms included cover: KNN and Naive Bayes plus a third completely new algorithm named Linear Classifier. The algorithms can work on real and discrete data and can be connected to a TDataSet descendant. They appropriately handle missing data and are all capable of incremental learning. The package includes a demo with reference results on standard domains, examples of usage and performance tests.
Product License: C
Source: On purchase/registration
File Status: Trial (work while IDE is running)
Compatible with: C2k10 C2k7 C2k9 CB64 CBXE CBXE2 CBXE3 CBXE4 CBXE5 CBXE6 CBXE7 D2007 D2009 D2010 DXE DXE2 DXE3 DXE4 DXE5 DXE6 DXE64 DXE7