Direkt zum Hauptbereich

The tensor

What is a tensor?

A  tensor is one structure which is needed and used from many AI environments.
Basically the tensor is a kind of a 4 dimensional matrix, containing parameters for 
dimensions and data itself. The 'data' has the information used  from the environment, 
this information may be color of a pixel, velocity of a motor, position of some object. 
Generally it is one or more junks of bytes. The order of this data contained in the tensor 
usually is a structure described through some matrices and vectors. 
From the software point of view, I am using tensor in my AI developments as a class 
described below

n is the number of samples. This is the size of the 

vector containing  the base data. For example, the 

base data is one picture, usually denoted by one array 

of 3, 4 pixel information (r, g, b, a). Sometimes named 

    as ‘channels’. Form the software point of view defined 

    as std::vector<Array of data..>

k is describing the number of the filters. Filters are 

    quadratic groups of data within one array of data. 

    For example one filter may be a quadratic small array 

    of our pixels (3x3), (4x4), (30x30),(...)

nr is the number of rows contained in only one array of 

    data, For examples 30 rows are there.

nc is the number of columns containing the data 

size is the total size of the Tensor. It may be calculated 

as the following product:

size = n x nr x nc  

For example n= 2, nr = 3, nc = 4 => size = 24