The SciPy issue scipy/scipy#21665 asked for tests of the derivative operator for dual numbers:
|
dual &operator/=(const dual &other) { |
|
for (size_t i = 0; i <= Order; ++i) { |
|
for (size_t j = 1; j <= i; ++j) { |
|
data[i] -= detail::fast_binom<T>(i - 1, j) * other.data[j] * data[i - j]; |
|
} |
|
|
|
data[i] /= other.data[0]; |
|
} |
|
|
|
return *this; |
|
} |
|
|
|
dual &operator/=(const value_type &other) { |
|
for (size_t i = 0; i <= Order; ++i) { |
|
data[i] /= other; |
|
} |
|
|
|
return *this; |
|
} |
At the time, there some doubts about the correctness of the implementation, which turned out to be unfounded. It was simply the case that the implementation is equivalent but not identical to the implementation from the reference paper Irwin had posted in ways that weren't immediately obvious to me, but perhaps should have been. It still seems like a good idea to add tests, but it will be more natural to add them to the C++ tests here.
The SciPy issue scipy/scipy#21665 asked for tests of the derivative operator for dual numbers:
xsf/include/xsf/dual.h
Lines 121 to 139 in d4217e1
At the time, there some doubts about the correctness of the implementation, which turned out to be unfounded. It was simply the case that the implementation is equivalent but not identical to the implementation from the reference paper Irwin had posted in ways that weren't immediately obvious to me, but perhaps should have been. It still seems like a good idea to add tests, but it will be more natural to add them to the C++ tests here.