arithmetic

The arithmetic expression to be executed. It has the following requirements:

1. Use {IDX} to refer to the No. IDX input tensor, where IDX is the index among the operands parameter.

2. The following arithmetic symbols are allowed: +, -, * (matrix multiplication), /(elementwise division), ^(power, the right operand to which must be a constant or 1x1 tensor), sin, cos, tan, asin, acos, atan, sinh, cosh, tanh and inv ( matrix inverse, for matrix division). Parentheses ( and ) are also supported.

3. Constants are allowed in the expression, but they cannot appear on both sides of the arithmetic symbols listed above, such as 5.0 + 1.0 is not valid, because the callers are required to pre-compute these arithmetic operations between constants. i.e., don't call the method with expression = {0} * (5.0 + 1.0), but {0} * 6.0 instead.

4. In terms of the computation priority ambiguity between trigonometry functions and power, we make the following rules: sin {0} ^ 2 means the sine of {0} ^ 2, whereas (sin {0}) ^ 2 means the power 2 of sine of operand 0.

Valid examples are:

• {0} + {1} * 2,

• sin (( {0} * 3.1415926 ) / 180),

• {0} * {1} note this is matrix multiplication,

• {0} * inv {1}.

the input operand to the arithmetic expression. The tensor indexed IDX in this array will replace the symbol {IDX} in expression. The size of this array must not exceed 10. All the tensors in this array must:

1. created use Tensor.MultiDimensionalInitInfo as initialization config, i.e., must be multi-dimensional tensor with no special usage;

2. has exactly two dimensions (current limitation, we will support multi-dimensional arithmetic operations shortly afterwards).

the tensor to store the result of the arithmetic operation.

If the usage is not allowed by the SpatialML run-time Framework, or if the framework encounter internal error and cannot perform the requested behavior.