2

I have a numpy 1dimensional array with n values lets call it xdata.

I would like to create a multidimensional numpy array lets call it xdataMulti such that each dimension of this array contains values in xdata that are in a certain range.

For example lets say that

xdata = np.array([-0.879645943,-0.7897614865,-0.7051130178,-0.6108652382,-0.5270894341,...])

and that I want to put all values of xdata between -0.9 and -0.6 in xdataMulti[:,0] and all values between -0.6 and 0 in xdataMulti[:,1], etc.

How do I create and fill xdataMulti (I know beforehand how many ranges I have and their endpoints but I dont know how many points of xdata are in each range, I have to loop through xdata to find out)?

Improve this question
3
  • 1
    That's not really what numpy (multi-)dimensional arrays are meant for! The length of the entries in each dimension should be the same. Commented Jan 25, 2017 at 22:03
  • 1
    You are describing a list of arrays of varying sizes. Commented Jan 25, 2017 at 22:10
  • Yes it is a list of arrays of varying sizes. I need each to be a numpy array in order to call a scipy function that only accepts numpy arrays. Commented Jan 25, 2017 at 22:18

2 Answers 2

Reset to default
2

If your number of ranges is not too large you can create them by hand

 gm09 = xdata >= -0.9
 gm06 = xdata >= -0.6
 g0   = xdata >= 0
 ranges = [xdata[gm09 & ~gm06], xdata[gm06 & ~g0]]
 result = [f(r) for r in ranges]

where f is your scipy function.

If your number of ranges is large, you can sort your data and then use searchsorted. Assuming you have the boundaries of blocks in the sorted array bnd:

xs = np.sort(xdata)
xbnd = np.searchsorted(xs, bnd)
ranges = [xs[l:r] for l, r in zip(xbnd[:-1], xbnd[1:])]
result = [f(r) for r in ranges]

Note that this differs from the first solution in that your ranges will be sorted.

If you need to preserve the original order you can use indirect sorting

order = np.argsort(xdata)
xbnd = np.searchsorted(xdata[order], bnd)
ordord = [np.sort(order[l:r]) for l, r in zip(xbnd[:-1], xbnd[1:])]
ranges = [xdata[oo] for oo in ordord]
result = [f(r) for r in ranges]
Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

In the middle code xs = np.sort(xdata) xbnd = np.searchsorted(xs, bnd) ranges = [xs[l:r] for l, r in zip(xbnd[:-1], xbnd[1:])] result = [f(r) for r in ranges] imagine xbnd is [6,18], then zip is [(6,18)] and the ranges object will only contain the range between 6 and 18 (between 0 and 6 and 18 and last are not picked).
@JennyToy Hm, isn't that how it should be? Having xbnd start at six means the smallest six values in xdata are below the smallest bin boundary, so there's no bin they should be in, at least in terms of the original Q. If you want to collect these outliers you can always do something like xbnd = np.r_[0, xbnd, len(xdata)]
1

multidimensional arrays must be rectangular, so if not all of your rows/columns have the same length you're out of luck.

However there are ways to deal with this. One way would be to create a list of arrays:

>>> arr = np.random.random((10))  # using this for values

>>> # each item in ranges corresponds to the lower and upper border.
>>> ranges = [(0, 0.3), (0.3, 0.6), (0.6, 1)]  

>>> [arr[(arr >= lower) & (arr < upper)] for lower, upper in ranges]
[array([ 0.15346374]),
 array([ 0.38144735,  0.45017858,  0.52710788,  0.36339812]),
 array([ 0.79770651,  0.77543868,  0.94824291,  0.87412998,  0.70890894])]

That's not a multidimensional array, just a list of arrays, but depending on your needs that may be enough.

Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.