Suppose I have a Numpy array of a bunch of coordinates [x, y]. I want to filter this array.
For all coordinates in the array with a same x-value, I want to keep only one coordinate:
The coordinate with the maximum for the y.
What is the most efficient or Pythonic way to do this. I will explain with an example below.

coord_arr= array([[10,5], [11,6], [12,6], [10,1], [11,0],[12,2]])

[10, 5] and [10,1] have the same x-value: x=10
maximum for y-values:
max(5,1) = 5
So I only keep coordinate [10,5]

Same procedure for x=11 and x=12
So I finally end up with:

filtered_coord_arr= array([[10,5],[11,6],[12,6]])

I have a solution by converting to a list and using list comprehension (see below).
But I am looking for a more efficient and elegant solution.
(The actual arrays are much larger than in this example.)
My solution:

coord_list = coord_arr.tolist()
x_set = set([coord[0] for coord in coord_list])
coord_max_y_list= []
for x in x_set:
    compare_list=[coord for coord in coord_list if coord[0]==x]
    coord_max = compare_list[compare_list.index(max([coord[1] for coord[1] in compare_list]))]
    coord_max_y_list.append(coord_max)
filtered_coord_arr= np.array(coord_max_y_list)

CodePudding user response：

you can refer below answer :

Solution :

coord_arr= np.array([[10, 5], [11, 6], [12, 6], [13,7], [10,1], [10,7],[12,2], [13,0]])

df = pd.DataFrame(coord_arr,columns=['a','b'])
df = df.groupby(['a']).agg({'b': ['max']})
df.columns = ['b']
df = df.reset_index()
filtered_coord_arr = np.array(df)

filtered_coord_arr

Output :

array([[10,  7],
       [11,  6],
       [12,  6],
       [13,  7]], dtype=int64)

CodePudding user response：

if your array in small you can just do it one line:

np.array([[x, max(coord[coord[:,0] == x][:,1])] for x in  set(coord[:,0])])

however that is not correct complexity, if array is big and you care about correct complexity , do like this:

d  = {}
for x, y in coord:
    d[x] = max(d.get(x, float('-Inf')), y)
np.array([[x, y] for x,y in d.items()])