Skip to content

Latest commit

 

History

History
234 lines (153 loc) · 5.72 KB

CHANGELOG.md

File metadata and controls

234 lines (153 loc) · 5.72 KB
Raw

v1.0.0

  • Improved DBSCAN implementation.
  • Added linear algebra.

v0.10.8

  • Added DBSCAN.
  • Added GetFrameTime method to StftInfo.
  • Added Copy extension method to Vec<T> and Mat<T>.
  • Updated MatFlat to 0.8.2.

v0.10.7

  • Added logistic regression.

v0.10.6

  • Added support for Range.

v0.10.5

  • Added real FFT.

v0.10.4

  • Added NMF.

v0.10.3

  • The length of the result of the Resample() extension method now matches MATLAB's implementation.
  • Fixed a potential issue where temporary matrix allocation, which is internally used in the library, might fail.
  • Added multiplicative update implementation for NMF.

v0.10.2

  • Fixed issue where resampling failed on long signal.

v0.10.1

  • Added MultivariateDistribution.Generate() method without args which uses Random.Shared.
  • Added resampling.

v0.10.0

  • Audio feature extraction supports complex spectrum.
  • Added random sampling method for distributions.
  • Some code cleanup.

v0.9.7

  • Added options for clustering algorithms.
  • Added options for ICA.
  • Some code cleanup.

v0.9.6

  • Fixed bug where SVD does not return if the matrix contains NaN values.
  • Added ICA.
  • Optimized vector value enumeration.
  • Some code cleanup.

v0.9.5

  • Improved Convolve method performance for short signals.
  • CsvFile now supports writing Vec<T> to a file.
  • Avoid memory allocation in enumeration.
  • Some code cleanup.

v0.9.4

  • Matrices can now be created using collection expressions.

v0.9.3

  • Added convolution.
  • Added (weighted) sum for scalars, vectors, and matrices.

v0.9.2

  • Added filter bank audio feature extraction.

v0.9.1

  • Removed unnecessary code that handle native dependencies.
  • Determinant() of matrix decomposition object for non-square matrices now throws an exception.

v0.9.0

  • Dropped native dependencies.

v0.8.1

  • Added NumFlat.IO.CsvFile class for CSV file IO.

v0.8.0

  • Added NumFlat.IO namespace for file IO.
  • Added NumFlat.IO.WaveFile class for wave file IO.

v0.7.7

  • Optimized k-means and GMM by omitting some calculations.
  • GMM now behaves the same as sklearn's default GMM.

v0.7.6

  • Added STFT and ISTFT.
  • Added WindowFunctions class.

v0.7.5

  • Moved FourierTransform class to SignalProcessing namespace.
  • Added the methods for framing and overlap-add.
  • Optimized Bhattacharyya distance.
  • Now UnsafeFastIndexer for vectors can be used with foreach.

v0.7.4

  • Now MathLinq supports weighted second order statistics for matrices.
  • Added the following extension methods.
    • double Skewness(this IEnumerable<double> xs, bool unbiased = true)
    • double Kurtosis(this IEnumerable<double> xs, bool unbiased = true)

v0.7.3

  • Now MathLinq supports weighted second order statistics for scalars (double and Complex).

v0.7.2

  • Now MathLinq supports second order statistics for scalars (double and Complex).

v0.7.1

  • Added GMM.

v0.7.0

  • Now Gaussian has a constructor that requires IEnumerable<Vec<double>>.
  • Added another overload of Svd.Decompose() that computes only S and U.
  • Now PCA uses SVD instead of EVD.
  • Added k-means clustering.

v0.6.2

  • Added MapInplace() to Vec<T> and Mat<T>.
  • Added Bhattacharyya() to Gaussian.
  • Added GetUnsafeFastIndexer() to Vec<T> and Mat<T>.

v0.6.1

  • Added FittingFailureException to indicate model fitting failure.
  • Added regularization parameter to ToGaussian().
  • Added pointwise scalar addition and subtraction for vectors and matrices.
  • Added Distance() to Vec<T>.
  • Added Mahalanobis() to Gaussian.

v0.6.0

  • Added the Gaussian and diagonal Gaussian distributions.
  • Optimized the Cholesky decomposition implementation.

v0.5.0

  • Added FFT.

v0.4.0

  • Improved error handling and error messages.
  • Added Mat<T>.InverseInplace().
  • Added PCA and LDA.

v0.3.2

  • Added the following in-place operations.
    • Vec<T>.ReverseInplace()
    • Vec<Complex>.ConjugateInplace()
    • Mat<T>.TransposeInplace()
    • Mat<Complex>.ConjugateInplace()
    • Mat<Complex>.ConjugateTransposeInplace()
  • Matrix decomposition methods now support Solve() against matrices as a set of RHS vectors.

v0.3.1

  • Added the weighted version of Mean, Variance, Covariance, StandardDeviation for vectors.

v0.3.0

  • Added the Determinant and LogDeterminant methods to matrix decomposition objects.
  • Now MathLinq supports Mean and Variance for matrices.
  • Optimized Covariance by utilizing the symmetry of matrices.
  • Added the following norm-related methods to Vec<T>.
    • Norm()
    • Norm(p)
    • L1Norm()
    • InfinityNorm()
    • Normalize()
    • Normalize(p)
  • Added the following norm-related methods to Mat<T>.
    • FrobeniusNorm()
    • L1Norm()
    • L2Norm()
    • InfinityNorm()
  • Added in-place operations for vectors and matrices.
    In-place operations have a method name with the suffix Inplace (NormalizeInplace for example).

v0.2.2

  • Added the following builder methods.
    • Vec<T> VectorBuilder.Fill<T>(int count, T value)
    • Vec<T> VectorBuilder.FromFunc<T>(int count, Func<int, T> func)
    • Mat<T> MatrixBuilder.Fill<T>(int rowCount, int colCount, T value)
    • Mat<T> MatrixBuilder.FromFunc<T>(int rowCount, int colCount, Func<int, int, T> func)
  • Added the Variance method to MathLinq.
  • Added the generalized eigen value decomposition.

v0.2.1

  • Revised the doc comments and error messages.
  • Added the eigen value decomposition.

v0.2.0

  • Revised the unit tests for better robustness.
  • Revised the doc comments and error messages.
  • Now the LU decomposition directly exposes L and U like the other decomposition methods.

v0.1.1

  • Revised the readme included in the NuGet package.

v0.1.0

  • This is the first release.