Unary Functions for Opus Expressions

There are a set of unary functions defined to use in expressions, as follows. These all operate on numpy arrays.

clip_to_zero

Returns the input values with all negative values clipped to 0.

exp

Returns an array consisting of $e$ raised to the input values.

ln

Returns an array of natural logarithms. Input values of 0 result in 0. (The intent is that this function be used on arrays of values, where 0 denotes a missing value. However, be cautious -- as you approach 0 from the positive side, the result value becomes more and more negative, and then suddenly returns to 0 at 0.)

ln_bounded

Returns an array of natural logarithms. Values less than 1 result in 0.

ln_shifted

Returns an array of natural logarithms. The input values are shifted by the second argument before taking the log. (The default shift value is 1.)

ln_shifted_auto

If the input values includes values that are less than or equal to 0, they are shifted so that that the minimum of the shifted values is 1 before taking the log. Otherwise the log is taken on the original values.

sqrt

Returns an array of square roots. Values less than 0 result in 0.

In addition, all of the functions in numpy are available. To avoid name collisions, the function name in an expression must include the package name numpy. For example, this expression gives you the reciprocals of all the values in a variable v:

numpy.reciprocal(v)