
Please add Matrix<T> broadcasting.
When working with Deep NNs, there are a lot of "Z = WX + b" operations going on, where matrix WX has the same number of rows as matrix b. But, matrix b is a single column matrix, so we can't just the built in add operations. With broadcasting, matrix b will be treated like a matrix with the same number of columns as WX, but where each column is identical.
1 vote 
Common Interface for Vector and Matrix
A common (gemeinsames) Interface for Vector and Matrix would be fine.
This would make it easier to implement common methods be using Generics.
Like:
T mapInplace<T>(T t) where T : IMatrix {
t.MapInplace(f => .... );
return f;}
1 vote 
Return Vector/Matrix on inplace methods
If someone does one of the inplace Methods, normally this object will be used immediately.
Therefore it makes sense (and is the common way) to return the inplace modified object.
So, instead of
public void MapInplace(Func<T, T> f, Zeros zeros = Zeros.AllowSkip);
a
public Matrix<T> MapInplace(Func<T, T> f, Zeros zeros = Zeros.AllowSkip);
would be good. Also for Vector, and all other inplace methods.
1 vote 
Support Weighted and Exponential Moving Average
Support Weighted and EMA (Exponential Moving Average) in addition to existing simple moving average in MathNet.Numerics.Statistics
1 vote 
Implement Laplace transforms
convert laplace space functions to time domain and viceversa
1 vote 
Implement an extension method ".ToVector()" for convert a IEnumberable to a Vector
if it is possible, an extension method ".ToMatrix()" would be also helpful
thank you~
1 vote 
1 vote

Improving sparse matrix storage CRS by allowing the user to provide the sparsity structure of the matrix.
I'm new to math.net, but I've found some issues with the libraries.
One of these is related to sparsematrix and sparsecompressedrowstorage
I've also wrote a post within the discussion forum, but I've recieved no response.
The problem is simple: I've tried desperately to find a way to assign directly the sparse matrix in CRS format, but I've failed.
It seems that the CSR must be constructed starting from an already existing matrix. This would imply a big performace issue and memory allocation problems.
For istance, if one want to write a program to solve some physical problem with finite element…1 vote 
Fix splines so that they can handle general lines
When adding the points of a circle ( Cos(t), sin(t) ) for t the values [0,6.28], I get complete garbage out.
It seems that the splines are unable to handle ambiguous values for y for a given x.
Splines should have a independent variable t, on which x and y depend. It seems this implementation has a t that somehow is coupled to x.
Interpolate should for example return a point, and the differentiate should return a vector for a given point.1 vote 
Multidimensional Arrays Support
Support multidimensional array and tensors with higher than 2D (i.e. Matrix). This would open Math.Net to more users who need this functionality daily.
3 votes 
Dense and sparse matrix slicing based on input user defined indices
This should allow functionality similar to MATLAB:
Suppose that A is the following matrixA = [ 10 20 30 40 ;
11 21 31 41;
12 22 32 42;
13 23 33 43]and I = { 1, 2, 4 } and J = { 3, 4 }
A_IJ = A[I, J]
should return a slice of the matrix according to the indices specified:
A_IJ = [ 30 40;31 41;
33 43 ]The same idea is very useful for arrays and it should work in both directions, for example:
A[I, J] += some3by2matrix;
The same…
1 vote 
Add trigonometrical functions for degrees
For most cases it can be overcome with multiplying pi/180 and using loops, but it would make code simpler. Coming from Octave/Matlab it is something I am used to have
2 votes 
Quasi Random Number Generation
Otherwise known as low discrepancy sequences:
https://en.m.wikipedia.org/wiki/Lowdiscrepancy_sequence1 vote 
Geometric Problems  Like Adjacent Points
Where's the namespace for geometric problems?
2 votes 
Implement the Studentized range distribution
https://en.wikipedia.org/wiki/Studentizedrangedistribution
This distribution is used in the Tukey test for post hoc analysis of an ANOVA test.
6 votes 
support infinite precision
I really love the infinite precision feature of Google's Android Calculator. If you enter a number like π, e, φ, √2, √3, √5, etc., you can scroll to the right to get an infinite number of digits.
Such calculations should be supported by Math.NET Numerics, where results can get queried to get more (decimal) digits. Something like an IEnumerable<DecimalDigit> interface.
4 votes 
Implement finding extrema for cubic splines
I wanted to find extrema for a spline and had to implement it by myself, using the CubicSpline.cs of MathNet.Numerics. Unfortunately the Class is constructes thus, that I could not access the spline parameters which would have allowed my to simply write an extension method.
I propose the following addition to the class "CubicSpline":
public enum eExtremePointType
{
Minimum, Maximum
}public class cExtremePointResult
{
public eExtremePointType Type { get; private set; }
public double x { get; private set; }public double y [get;private set;}
public cExtremePointResult(double in_x, double in_y, eExtremePointType in_type)
{x=inx;
y=iny;
Type=in_type;
…}
3 votes 
Eiganvalues and Eiganvectors
A function that can evaluate a square matrix for eiganvalues and eiganvectors similar to MATLAB's eig() function.
1 vote 
Implement Selected MATLAB Core Functionality
Implement a new class which is intended to provide syntax matched core MATLAB functionalities, such as flip(), linspace(), etc. I know linspace() and logspace() are already present with reordered inputs, but matching MATLAB syntax would prevent a lot of rewriting. Some of these could simply be wrappers for existing Math.NET functions, just so syntax matches.
Validation of this functionality would be such that inputs/outputs match the MATLAB implementation. Simple functions at the start, and more complex ones as time goes on.
3 votes 
Valuation of top eigenvalues/eigenvectors only
Hi,
CPU time to perform Matrix diagonalization scales like n^3 where n is the size of the matrix. In many applications, however, one needs to know the top eigenvalues/eigenvectors only.
MKL provides this: "LAPACKE_dsyevr" where one can select how many eigenvalues one wants to compute.
Would be great if this functionality was available in mathnet!1 vote
 Don't see your idea?