Closed
Description
For v0.2 I'm dropping support for ChebHash. I anticipate that MonteCarloHash
will be the exclusive family of hashes for function spaces.
There are multiple reasons to remove ChebHash
:
- Compared to
MonteCarloHash
,ChebHash
is very limited in its usefulness. It's much more difficult to extend to multidimensional spaces or to irregular (i.e., non-cubic) domains. As a result, it's mostly of theoretical interest, and has limited practical use. - It's a lot more difficult to get bounds on the effectiveness of
ChebHash
thanMonteCarloHash
, primarily because the effectiveness ofChebHash
is highly dependent on the degree of Chebyshev polynomial you use to approximate the input function. - Generally speaking,
ChebHash
is more complex to maintain long-termMonteCarloHash
, especially if we want to find an optimal number of Chebyshev coefficients for a provided input function. ChebHash
requires theFFTW
dependency, which is otherwise completely unused by the package.
It'd be a better use of time to stop focusing on ChebHash
, and to instead focus work on MonteCarloHash
for function hashing.