Update Backend for convolution layer

[pg020196]

Name	Marker	Type
KERNEL_VALUES	###kernelValues###	const float array with length of ###indicesKernelValues###
KERNEL_START_INDEX	###kernelvaluesstartindex###	const uint16_t array with length of ###dimNumberLayers###
KERNEL_WIDTH	###kernelwidth###	const uint16_t array with length of ###dimNumberLayers###
KERNEL_HEIGHT	###kernelheight###	const uint16_t array with length of ###dimNumberLayers###
HORIZONTAL_DILATION_RATE	###HorizontalDilationRate###	const uint16_t array with length of ###dimNumberLayers###
HORIZONTAL_DILATION_RATE	###VerticalDilationRate###	const uint16_t array with length of ###dimNumberLayers###

Furthermore, entries in the following existing arrays must be added: activation_function bias_enabled horizontal_stride vertical_stride

We also must somehow specify whether a zero padding or a value padding should be applied, e.g. through a padding type field with the following assignment: zero padding: 0 value padding: 1

[ChristophBrandl]

@pg020196 Additional tasks/questions:

• Where comes the information for zero/value padding field from?
• Formula for correct calculation of output-size of convolution layer
• Where should the bias values for convolution layer be stored?

[pg020196]

• I think for now we can assume that only the convolution layer uses value padding
• isn’t it the same formula as for pooling?
• what do you think of the usual bias array

[pg020196]

Due to the project time frame, the project team decided not to implement the convolution layer.