CODEWITHSUNDEEP
pythonperformancenumpyvectorization
Jonno
Jonno

Reputation: 787

Find where array values increase monotonically over some value

I'm trying to find the locations in an array where the values increase monotonically such that the total change in value is greater than k. Ie for k = 5 and data = [1, 4, 5, 7, 10, 9, 6, 14, 3, 4] I would want to return:

[4, 7]

In general the array would be floats.

Edit to further clarify: In the example the data values increase monotonically by greater than 5 in two intervals. The first is from element 0 to 4 (where the data values increase by 9), the 2nd is from element 6 to 7 (where the data values increase by 8). Thus, I want to report the end of each valid interval.

I want to avoid loops and make this efficient. If necessary the range of consecutive values to check could be limited.

Upvotes: 0

Views: 871

Answers (1)

Divakar
Divakar

Reputation: 221624

Here's a vectorized approach -

d = a[1:] - a[:-1]
mask = np.concatenate(( [False], d > 0, [False] ))
start = np.flatnonzero(mask[1:] > mask[:-1])
stop = np.flatnonzero(mask[1:] < mask[:-1])
count = np.bincount(np.repeat(np.arange(start.size) ,stop - start), d[mask[1:-1]])
out = stop[count > k]

Sample input, output -

In [202]: a # Adding in two monotonically decreasing elems at start for variety
Out[202]: array([10,  7,  1,  4,  5,  7, 10,  9,  6, 14,  3,  4])

In [203]: k = 5

In [204]: out
Out[204]: array([6, 9])

Upvotes: 2

Related Questions