multidimensional kerneldiff and butterdiff#181
Conversation
…xis parameter and support multidimensional data
| [(1, 0), (2, 2), (1, 0), (2, 2)], | ||
| [(1, 0), (3, 3), (1, 0), (3, 3)]], | ||
| spectraldiff: [[(-15, -15), (-14, -15), (0, -1), (0, 0)], | ||
| spectraldiff: [[(-15, -15), (-14, -14), (0, -1), (0, 0)], |
There was a problem hiding this comment.
spectraldiff uses convolution, which is now based on scipy's convolve1d rather than numpy's convolve, which changes its answers ever so slightly.
| for i in range(num_iterations): | ||
| x_padded = np.pad(x_hat, pad_width, mode='symmetric') # pad with repetition of the edges | ||
| x_hat = np.convolve(x_padded, kernel, 'valid')[:len(x)] # 'valid' slices out only full-overlap spots | ||
| x_hat = convolve1d(x_hat, kernel, axis=axis, mode='reflect') # 'reflect' pads the signal with repeats |
There was a problem hiding this comment.
This function is great. No more need to pad things ourselves nor carefully slice out the answer to ensure same-length.
|
|
||
| # included code | ||
| from pynumdiff.finite_difference import second_order as finite_difference | ||
| from pynumdiff.finite_difference import finitediff |
There was a problem hiding this comment.
Why I wasn't doing this before I'm not sure.
| x_hat = x | ||
| for _ in range(num_iterations): | ||
| x_hat = scipy.signal.medfilt(x_hat, window_size) | ||
| x_hat = scipy.signal.medfilt(x_hat, s) |
There was a problem hiding this comment.
medfilt has a slightly different mechanism for working with multidimensional data; you have to give a multidimensional kernel. Where that kernel has dimensional length 1, no filtering occurs.
| ``` | ||
|
|
||
| where `x` is data, `dt` is a step size, and various keyword arguments control the behavior. Some methods support variable step size, in which case the second parameter is renamed `dt_or_t` and can receive either a constant step size or an array of values to denote sample locations. | ||
| where `x` is data, `dt` is a step size, and various keyword arguments control the behavior. Some methods support variable step size, in which case the second parameter is renamed `dt_or_t` and can receive either a constant step size or an array of values to denote sample locations. Some methods support multidimensional data, in which case there is an `axis` argument to control the dimension differentiated along. |
There was a problem hiding this comment.
Calling attention to the fact we support multidimensional data.
1 participant
extended the non-legacy smooth_finite_difference methods to have an axis parameter and support multidimensional data