Math.NET Numerics Activity

Created Feature: Cache Oblivious Matrix Multiplication [5641]

jvangael — Wed, 28 Jul 2010 14:59:34 -0700

Experiment whether it buys us anything to have cache oblivious matrix operations for the managed backend.

http://www.compgeom.com/co-chap/chap.pdf
http://www-math.mit.edu/~stevenj/18.335/oblivious-matmul.pdf

New Post: How can I calculate Eigenvalues/vectors?

jvangael — Wed, 28 Jul 2010 03:13:35 -0700

Hey Legoangthang,

We are currently porting the Complex matrices to Math.Net Numerics. Before that's done we won't be able to expose the eigenvalue creation. Hopefully we get this done before the end of the summer though!

Cheers, Jurgen

New Post: How can I calculate Eigenvalues/vectors?

lehoangthanh — Mon, 26 Jul 2010 09:03:24 -0700

I read the "Code Samples Documentation" about using "Linear Algebra class" to find eigenvectors/values. However, when I used this code, it caused some errors. How can I solve? This is the link of code samples: http://numerics.mathdotnet.com/doc/%28X%281%29S%28kvsekp45nm141teol3edgo3k%29%29/Default.aspx?Page=CodeSamples&NS=&AspxAutoDetectCookieSupport=1

New Post: Performance Helper Idea

cuda — Mon, 26 Jul 2010 05:59:13 -0700

Hi Jurgen, That would be helpful, but we should probably put it in the next release. Marcus

Added sparse interface to the F# front-end.

Jurgen Van Gael — Sun, 25 Jul 2010 07:21:37 -0700

m src/FSharp/DenseMatrix.fs
m src/FSharp/DenseVector.fs
m src/FSharp/FSharp.fsproj
m src/FSharp/Main.fs
m src/FSharp/Matrix.fs
+ src/FSharp/SparseMatrix.fs
+ src/FSharp/SparseVector.fs
m src/FSharp/Vector.fs
m src/FSharpUnitTests/Program.fs
m src/Numerics/LinearAlgebra/Double/SparseVector.cs
m src/UnitTests/LinearAlgebraTests/Double/SparseVectorTest.cs Added sparse interface to the F# front-end.
Fixed unit tests for new functional friendly vector interface.

Updated unit tests to reflect new Vector API.

Jurgen Van Gael — Sun, 25 Jul 2010 07:21:36 -0700

m src/Numerics/LinearAlgebra/Double/DenseVector.cs
m src/Numerics/LinearAlgebra/Double/SparseVector.cs
m src/UnitTests/LinearAlgebraTests/Double/VectorTests.Arithmetic.cs Updated unit tests to reflect new Vector API.

Bug fixes in the unit tests for the new functional friendly API.

Jurgen Van Gael — Sun, 25 Jul 2010 07:21:36 -0700

m src/Numerics/LinearAlgebra/Double/DenseVector.cs
m src/Numerics/LinearAlgebra/Double/SparseVector.cs
m src/Numerics/LinearAlgebra/Double/Vector.cs
m src/UnitTests/LinearAlgebraTests/Double/VectorTests.Arithmetic.cs Bug fixes in the unit tests for the new functional friendly API.
Typo fix: Sparce -> Sparse

Fixed all bugs on Vector and DenseVector for the new functional friendly interface.

Jurgen Van Gael — Sun, 25 Jul 2010 07:21:36 -0700

m src/Numerics/LinearAlgebra/Double/DenseVector.cs
m src/Numerics/LinearAlgebra/Double/Vector.cs
m src/UnitTests/LinearAlgebraTests/Double/DenseVectorTests.cs Fixed all bugs on Vector and DenseVector for the new functional friendly interface.

Changed the Vector API to make methods more functional programming friendly.

Jurgen Van Gael — Sun, 25 Jul 2010 07:21:35 -0700

m src/Numerics/LinearAlgebra/Double/DenseVector.cs
m src/Numerics/LinearAlgebra/Double/SparseVector.cs
m src/Numerics/LinearAlgebra/Double/Vector.cs Changed the Vector API to make methods more functional programming friendly.

New Post: Performance Helper Idea

jvangael — Sun, 25 Jul 2010 09:56:20 -0700

Hi Guys,

While working on the sparse vector stuff I started wondering about the following: would it be worth adding a very lightweight logging facility to our code (perhaps using a separate build) where we can notify the user that he might be doing something potentially dangerous (numerically unstable, poor performance): e.g. adding a scalar to a sparse vector "SparseVector.Add(3.0)". In "tracing" mode, we can help the user find performance bottlenecks.

Probably not for v1, but perhaps in our next iteration?

New Post: Vector.PointwiseMultiply/Subtract/Multiply/...

jvangael — Tue, 20 Jul 2010 14:07:43 -0700

OK brilliant; I'll go through the code and make the changes. Essentially all Vector methods which take a Vector result as a parameter I will leave as is, their equivalents without that parameter I will change the return parameter from void to Vector. Cheers, J

New Post: Vector.PointwiseMultiply/Subtract/Multiply/...

cuda — Mon, 19 Jul 2010 17:19:17 -0700

Hi Jurgen,

Duh! Right, we can optimize w.Subtract(v, w) to not do the unnecessary copy - so lets not add the InplaceXXX methods.

Regards,

Marcus

New Post: Vector.PointwiseMultiply/Subtract/Multiply/...

jvangael — Mon, 19 Jul 2010 13:40:39 -0700

Hi Marcus, I agree that we should keep the inplace methods for efficiency reasons. However, at the moment you can write w.Subtract(v, w) to make sure you overwrite the current vector (we can internally optimize this call to do the same InplaceXxx. This is a bit shorter than writing w.InplaceSubtract(v) Any toughts? Cheers, J

New Post: Vector.PointwiseMultiply/Subtract/Multiply/...

cuda — Tue, 13 Jul 2010 00:46:23 -0700

Hi Jurgen,

I think we should keep the inplace methods on both the vector and matrix classes. It reduces overhead and memory requirements when dealing with large amounts of data. We could rename them to InplaceXXX(...) and then and Vector/Matrix XXX(...). Would that work?

Regards,

Marcus

New Post: Vector.PointwiseMultiply/Subtract/Multiply/...

jvangael — Tue, 13 Jul 2010 00:34:43 -0700

Hi Guys,

Another API suggestion I want to make; all our Vector methods like PointwiseMultiply,Subtract,Multiply come in two flavors: e.g.

void Subtract(Vector v)

void Subtract(Vector v, Vector result)

In other words, if you want to do an in-place subtract, you can either call Subtract(v) or Subtract(v,v). My suggestion would be that we turn the "void Subtract(v)" into a "Vector Subtract(v)" which is immutable but returns a newly created vector that is the subtraction of "this" and "v". First, if you want to do inplace stuff, calling Subtract(v,v) is not really any overhead at all wheras if you want a new vector, you'd have to write

var wcopy = w.Clone()

w.Subtract(v, wcopy)

which is a bit of a hassle (I find) and completely breaks the flow of functional programs in F#.

In other words, my suggestion would be to make sure our API has both mutable and immutable methods.

Any thoughts,

Jurgen

Reformatting of Vector.Norm and Vector.Normalize code.

Jurgen Van Gael — Mon, 12 Jul 2010 08:34:32 -0700

m src/Numerics/LinearAlgebra/Double/DenseVector.cs
m src/Numerics/LinearAlgebra/Double/Matrix.Arithmetic.cs
m src/Numerics/LinearAlgebra/Double/SparseVector.cs
m src/Numerics/LinearAlgebra/Double/Vector.cs
m src/UnitTests/LinearAlgebraTests/Double/MatrixTests.Arithmetic.cs
m src/UnitTests/LinearAlgebraTests/Double/VectorTests.Norm.cs Reformatting of Vector.Norm and Vector.Normalize code.

New Post: Vector Norm/Normalize

cuda — Mon, 12 Jul 2010 00:43:36 -0700

Hi Jurgen, Makes sense. Marcus

New Post: Vector Norm/Normalize

jvangael — Mon, 12 Jul 2010 00:33:30 -0700

Hi Guys,

I'd like to slightly simplify the vector API if everyone's ok with this. I'd like to reduce the functions to

Vector.Norm(int p)

Vector.NormInfinity()

Vector.Normalize(int p)

rather than have Norm, SquaredNorm, Norm1, NormP, NormInfinity, and Normalize (where the last one always does 2-norm). For Norm, I think having 5 methods is a bit overkill. Particularly for Normalize I think we should parameterize as we can't Normalize in 1-norm atm. Is everyone OK if I make those changes?

words added to the dictionary

Marcus Cuda — Sat, 10 Jul 2010 00:47:57 -0700

m src/MathNet.Numerics.5.0.ReSharper words added to the dictionary

style cop fix

Marcus Cuda — Sat, 10 Jul 2010 00:47:53 -0700

m src/Numerics/LinearAlgebra/Double/Matrix.cs style cop fix

io: added matlab reader

Marcus Cuda — Sat, 10 Jul 2010 00:47:46 -0700

+ data/Matlab/A.mat
+ data/Matlab/collection-nocompress.mat
+ data/Matlab/collection.mat
+ data/Matlab/matrices.txt
+ data/Matlab/sparse-large.mat
+ data/Matlab/sparse-small.mat
+ data/Matlab/v.mat
+ data/Matrices/bp___200.mtx
+ data/Matrices/can24_pattern_symmetric_coordinate_24.mtx
+ data/Matrices/gear_integer_general_coordinate_100.mtx
+ data/Matrices/hilbert_real_symmetric_array_10.mtx
+ data/Matrices/random_real_general_array_100.mtx
+ data/Matrices/random_real_general_array_10_20.mtx
+ data/Matrices/random_real_general_array_20_10.mtx
+ data/Matrices/random_real_general_coordinate_10_20.mtx
+ data/Matrices/random_real_general_dense_10_20.comma
+ data/Matrices/random_real_general_dense_10_20.space
+ data/Matrices/random_real_general_dense_10_20.tab
+ src/Numerics/LinearAlgebra/Common/IO/Matlab/ArrayClass.cs
+ src/Numerics/LinearAlgebra/Common/IO/Matlab/ArrayFlags.cs
+ src/Numerics/LinearAlgebra/Common/IO/Matlab/DataType.cs
+ src/Numerics/LinearAlgebra/Double/IO/Matlab/MatlabFile.cs
+ src/Numerics/LinearAlgebra/Double/IO/Matlab/MatlabParser.cs
+ src/Numerics/LinearAlgebra/Double/IO/MatlabReader.cs
m src/Numerics/LinearAlgebra/Double/Matrix.cs
m src/Numerics/Numerics.csproj
m src/Numerics/Properties/Resources.Designer.cs
m src/Numerics/Properties/Resources.resx
+ src/UnitTests/LinearAlgebraTests/Double/IO/MatlabReaderTests.cs
m src/UnitTests/LinearAlgebraTests/Double/MatrixTests.cs
m src/UnitTests/UnitTests.csproj io: added matlab reader matrix: added general version of matrix norms - need to add dense and sparse specific

io: ported delimited writer. added support for different cultures and column headers

Marcus Cuda — Sat, 10 Jul 2010 00:44:44 -0700

+ src/Numerics/LinearAlgebra/Double/IO/DelimitedWriter.cs
+ src/Numerics/LinearAlgebra/Double/IO/MatrixWriter.cs
m src/Numerics/Numerics.csproj
m src/Silverlight/Silverlight.csproj
m src/UnitTests/LinearAlgebraTests/Double/IO/DelimitedReaderTests.cs
+ src/UnitTests/LinearAlgebraTests/Double/IO/DelimitedWriterTests.cs
m src/UnitTests/UnitTests.csproj io: ported delimited writer. added support for different cultures and column headers

Source code checked in, #4a00c3114a53

Marcus Cuda — Sat, 10 Jul 2010 07:44:00 -0700

words added to the dictionary

Source code checked in, #d11dcecd45e5

Marcus Cuda — Sat, 10 Jul 2010 07:28:51 -0700

style cop fix

Source code checked in, #6569d60a6f5e

Marcus Cuda — Sat, 10 Jul 2010 07:23:46 -0700

io: added matlab reader matrix: added general version of matrix norms - need to add dense and sparse specific

Edited Feature: Port Matlab parser [5638]

cuda — Sat, 10 Jul 2010 05:37:18 -0700

Port the dnA matlab parser.

io: ported delimited reader. made it generic, added support for different cultures, and fixed regex bug

Marcus Cuda — Fri, 09 Jul 2010 21:32:47 -0700

+ src/Numerics/LinearAlgebra/Double/IO/DelimitedReader.cs
+ src/Numerics/LinearAlgebra/Double/IO/MatrixReader.cs
m src/Numerics/Numerics.csproj
m src/Numerics/Properties/Resources.Designer.cs
m src/Numerics/Properties/Resources.resx
m src/Silverlight/Silverlight.csproj
+ src/UnitTests/LinearAlgebraTests/Double/IO/DelimitedReaderTests.cs
m src/UnitTests/UnitTests.csproj io: ported delimited reader. made it generic, added support for different cultures, and fixed regex bug

io: moving io code into main project

Marcus Cuda — Fri, 09 Jul 2010 21:32:46 -0700

- src/IO/IO.csproj
- src/IO/Properties/AssemblyInfo.cs io: moving io code into main project

added io directory

Marcus Cuda — Fri, 09 Jul 2010 21:32:46 -0700

m src/MathNet.Numerics.sln
m src/Numerics/Numerics.csproj
m src/Silverlight/Silverlight.csproj
m src/UnitTests/UnitTests.csproj added io directory

separate Version.cs T4 script for silverlight project (alternative would be to link directly to Numerics' Version.cs file)

Christoph Ruegg — Fri, 09 Jul 2010 09:40:36 -0700

m src/Numerics/Version.cs
m src/Silverlight/Silverlight.csproj
+ src/Silverlight/Version.cs
+ src/Silverlight/Version.tt separate Version.cs T4 script for silverlight project (alternative would be to link directly to Numerics' Version.cs file)

factor: Merged Andriy's factorization code for user defined matrices

Marcus Cuda — Fri, 09 Jul 2010 00:55:44 -0700

m src/MathNet.Numerics.5.0.ReSharper
m src/Numerics/Algorithms/LinearAlgebra/ManagedLinearAlgebraProvider.cs
m src/Numerics/LinearAlgebra/Double/Factorization/Cholesky.cs
m src/Numerics/LinearAlgebra/Double/Factorization/DenseCholesky.cs
m src/Numerics/LinearAlgebra/Double/Factorization/LU.cs
m src/Numerics/LinearAlgebra/Double/Factorization/QR.cs
m src/Numerics/LinearAlgebra/Double/Factorization/Svd.cs
+ src/Numerics/LinearAlgebra/Double/Factorization/UserCholesky.cs
+ src/Numerics/LinearAlgebra/Double/Factorization/UserLU.cs
+ src/Numerics/LinearAlgebra/Double/Factorization/UserQR.cs
+ src/Numerics/LinearAlgebra/Double/Factorization/UserSvd.cs
m src/Numerics/Numerics.csproj
m src/Silverlight/Silverlight.csproj
m src/UnitTests/LinearAlgebraTests/Double/Factorization/CholeskyTests.cs
m src/UnitTests/LinearAlgebraTests/Double/Factorization/LUTests.cs
m src/UnitTests/LinearAlgebraTests/Double/Factorization/QRTests.cs
m src/UnitTests/LinearAlgebraTests/Double/Factorization/SvdTests.cs
+ src/UnitTests/LinearAlgebraTests/Double/Factorization/UserCholeskyTests.cs
+ src/UnitTests/LinearAlgebraTests/Double/Factorization/UserLUTests.cs
+ src/UnitTests/LinearAlgebraTests/Double/Factorization/UserQRTests.cs
+ src/UnitTests/LinearAlgebraTests/Double/Factorization/UserSvdTests.cs
m src/UnitTests/LinearAlgebraTests/Double/MatrixLoader.cs
m src/UnitTests/LinearAlgebraTests/Double/UserDefinedMatrixTests.cs
m src/UnitTests/UnitTests.csproj factor: Merged Andriy's factorization code for user defined matrices

Source code checked in, #78e8b9af1ac8

Marcus Cuda — Fri, 09 Jul 2010 07:55:05 -0700

factor: Merged Andriy's factorization code for user defined matrices

New Post: When will Math.Net Numerics be available?

cuda — Wed, 07 Jul 2010 23:52:34 -0700

We are planning for a beta release in September and a final release before the end of the year. You can always grab the latest code from GitHub or CodePlex (we are now synchronizing the CodePlex repo with GitHub).

Regards,

Marcus

New Post: When will Math.Net Numerics be available?

benjaminarai — Wed, 07 Jul 2010 16:54:18 -0700

Is this still the case? Will there be an official release?

lu: merged Andriy's LU additions

Marcus Cuda — Tue, 06 Jul 2010 04:44:42 -0700

m src/Numerics/Algorithms/LinearAlgebra/Atlas/AtlasLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/ILinearAlgebraProviderOfT.cs
m src/Numerics/Algorithms/LinearAlgebra/ManagedLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/Mkl/MklLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/NativeAlgebraProvider.include
m src/Numerics/LinearAlgebra/Double/Factorization/DenseLU.cs
m src/Numerics/LinearAlgebra/Double/Factorization/LU.cs
m src/Numerics/LinearAlgebra/Double/Matrix.Arithmetic.cs
m src/UnitTests/LinearAlgebraTests/Double/Factorization/LUTests.cs lu: merged Andriy's LU additions

Implemented Matrix Rank, ConditionNumber, Determinant, Inverse.

abratiychuk — Tue, 06 Jul 2010 04:44:41 -0700

m src/Numerics/Algorithms/LinearAlgebra/Atlas/AtlasLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/ILinearAlgebraProviderOfT.cs
m src/Numerics/Algorithms/LinearAlgebra/ManagedLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/Mkl/MklLinearAlgebraProvider.cs
m src/Numerics/LinearAlgebra/Double/Factorization/DenseLU.cs
m src/Numerics/LinearAlgebra/Double/Factorization/LU.cs
m src/Numerics/LinearAlgebra/Double/Matrix.Arithmetic.cs
m src/Numerics/LinearAlgebra/Double/Matrix.cs
m src/UnitTests/LinearAlgebraTests/Double/Factorization/LUTests.cs Implemented Matrix Rank, ConditionNumber, Determinant, Inverse.
Finished LU in ManagedLinearAlgebraProvider

MathNetNumerics: Dense QR and SVD have been added to the repo - sparse versions coming soon. #numerics #math

Sun, 04 Jul 2010 07:11:37 -0700

MathNetNumerics: Dense QR and SVD have been added to the repo - sparse versions coming soon. #numerics #math

added missing commit

Marcus Cuda — Sat, 03 Jul 2010 23:49:39 -0700

m src/MathNet.Numerics.5.0.ReSharper
m src/Numerics/Algorithms/LinearAlgebra/ManagedLinearAlgebraProvider.cs
m src/Numerics/Properties/Resources.resx added missing commit

Source code checked in, #d3c883cb326f

Marcus Cuda — Sun, 04 Jul 2010 06:49:21 -0700

added missing commit

tweaked work[0] return value

Marcus Cuda — Sat, 03 Jul 2010 23:43:14 -0700

m src/Numerics/Algorithms/LinearAlgebra/ManagedLinearAlgebraProvider.cs
m src/Numerics/Properties/Resources.Designer.cs tweaked work[0] return value

svd: merged Andriy's svd code and qr work array update

Marcus Cuda — Sat, 03 Jul 2010 23:23:20 -0700

m src/Numerics/Algorithms/LinearAlgebra/Atlas/AtlasLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/ILinearAlgebraProviderOfT.cs
m src/Numerics/Algorithms/LinearAlgebra/ManagedLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/Mkl/MklLinearAlgebraProvider.cs
m src/Numerics/Algorithms/LinearAlgebra/NativeAlgebraProvider.include
m src/Numerics/LinearAlgebra/Double/Factorization/DenseQR.cs
+ src/Numerics/LinearAlgebra/Double/Factorization/DenseSvd.cs
m src/Numerics/LinearAlgebra/Double/Factorization/ExtensionMethods.cs
m src/Numerics/LinearAlgebra/Double/Factorization/QR.cs
+ src/Numerics/LinearAlgebra/Double/Factorization/Svd.cs
m src/Numerics/Numerics.csproj
m src/Numerics/Properties/Resources.Designer.cs
m src/Numerics/Properties/Resources.resx
m src/Silverlight/Silverlight.csproj
+ src/UnitTests/LinearAlgebraTests/Double/Factorization/SvdTests.cs
m src/UnitTests/UnitTests.csproj svd: merged Andriy's svd code and qr work array update

MathNetNumerics: Good to see more and more support for numerical computing on .NET - NumPy & SciPy supported by Microsoft http://bit.ly/bdRa7u

Sun, 04 Jul 2010 01:10:27 -0700

MathNetNumerics: Good to see more and more support for numerical computing on .NET - NumPy & SciPy supported by Microsoft http://bit.ly/bdRa7u

MathNetNumerics: initial sparse vector and matrix implementations have been added - thanks Andriy! next up QR, SVD, and Eigen factorizations #numerics #math

Sun, 27 Jun 2010 01:02:06 -0700

MathNetNumerics: initial sparse vector and matrix implementations have been added - thanks Andriy! next up QR, SVD, and Eigen factorizations #numerics #math

MathNetNumerics: Lost in Math.NET Codenames? http://bit.ly/aSZeZt

Mon, 26 Apr 2010 15:02:36 -0700

MathNetNumerics: Lost in Math.NET Codenames? http://bit.ly/aSZeZt

MathNetNumerics: RT @cuda: Converting Math.NET Numerics to .NET 4 - we'll now use the built in Complex and Parallel classes.

Sun, 25 Apr 2010 10:41:33 -0700

MathNetNumerics: RT @cuda: Converting Math.NET Numerics to .NET 4 - we'll now use the built in Complex and Parallel classes.

MathNetNumerics: Thanks to @cuda #mathnet #numerics now has nightly dev builds!

Fri, 29 Jan 2010 06:09:15 -0800

MathNetNumerics: Thanks to @cuda #mathnet #numerics now has nightly dev builds!

MathNetNumerics: 2010.1.28.762 Build is up on CodePlex - http://bit.ly/9rRjHw #mathnet #numerics

Thu, 28 Jan 2010 02:55:28 -0800

MathNetNumerics: 2010.1.28.762 Build is up on CodePlex - http://bit.ly/9rRjHw #mathnet #numerics

MathNetNumerics: 2010.1.27.699 Build is up on CodePlex - http://bit.ly/9rRjHw #mathnet #numerics

Wed, 27 Jan 2010 01:48:38 -0800

MathNetNumerics: 2010.1.27.699 Build is up on CodePlex - http://bit.ly/9rRjHw #mathnet #numerics

MathNetNumerics: Math.Net has matrix addition and subtraction again. Hope to have time over the holidays to have a minimal matrix implementation again!

Thu, 17 Dec 2009 15:07:33 -0800

MathNetNumerics: Math.Net has matrix addition and subtraction again. Hope to have time over the holidays to have a minimal matrix implementation again!

MathNetNumerics: Old Math.NET blog/forum/tracker is up and running again, sorry for the inconvenience.

Fri, 11 Dec 2009 13:47:17 -0800

MathNetNumerics: Old Math.NET blog/forum/tracker is up and running again, sorry for the inconvenience.

MathNetNumerics: Sorry for the Math.NET server issues (old forum, old tracker). Please use the codeplex-based forum/tracker instead, see http://bit.ly/6QhBmn

Thu, 10 Dec 2009 01:22:39 -0800

MathNetNumerics: Sorry for the Math.NET server issues (old forum, old tracker). Please use the codeplex-based forum/tracker instead, see http://bit.ly/6QhBmn

MathNetNumerics: Math.NET Numerics now builds under #Silverlight 3, thanks @cuda

Tue, 01 Dec 2009 04:35:22 -0800

MathNetNumerics: Math.NET Numerics now builds under #Silverlight 3, thanks @cuda