implement numpy Matrix Class

[dotChris90]

following the API doc https://docs.scipy.org/doc/numpy-1.15.1/reference/generated/numpy.matrix.html

[dotChris90]

okay understood. Just one question in general here.

In Python there is the possibility to create an array of lists containing doubles for a matrix. Also there is the possibility to create array of arrays. For sure the array of array strategy brings a better performance when using matrix operations.

In C# we would have possibility of NDArray of NArrays of doubles or NDArray of usual double array. I am very sure, that NDArray of NDArray will lead to some performance issues since NDArray is an adapter class of a true array. But in case of NDArray<double[]> we could get full performance support.

So final question : shouldn't we design NumSharp in a way that we avoid nested NDArrays?

[Oceania2018]

We should consider performance first. Let's make a big change, like separate NDArray as 1 dim array, NDArray2 as 2 dim array, and fore T as value type.

if(!(typeof(T).equals(typeof(int32))) && !(typeof(T).equals(typeof(decimal))))

[Oceania2018]

Please retire the Matrix class. If you insist to keep it, I suggest that Matrix inherits NDArray<T>, just keep Shape = new List<int>() {0, 0}, the only different is Shape property. I reopen this issue till @dotChris90 fix it. Thanks.

[dotChris90]

Agree

[dotChris90]

Merged. Only inv I removed because reimplement in ndarray.

[dotChris90]

remove because matrix now inheritage from NDArray