cdist | usable configuration management | Configuration Management library

I have two numpy arrays: ones is some coordinates (x, y, z) with an id and another one is only some coordinates (x, y, z):

I would like to build a function npbatch(U,X) which compares data points in an input matrix (U) with data points in a training matrix (X) and gets me the index of X with the shortest euclidean distance to the data point in U. I would like to avoid any loops to increase the performance and I would like to use the function scipy.spatial.distance.cdist to compute the distance.

Example Input:

I want to quickly (say, <1 ms) perform a nearest neighbor search by quadrant.

Input of each search:

• a fixed set of points in 2D space (black points in image). This remains constant across searches, so it can be stored in an efficient data structure. Number of points between 1 and 1000.
• a point whose location is different for each search (red point in image). Abstractly, this divides the 2D space into 4 quadrants (separated by red lines in image) much like the origin in a Cartesian coordinates.

Output of each search:

• black point from each red quadrant that is closest (circled blue in image) to the red point.

The output should usually be 4 points, one from each quadrant. But there are some possible edge cases:

• a quadrant to have no black points; collect no points for this quadrant
• a quadrant has multiple ties; collect all such points for this quadrant
• same point lies on boundaries (red lines) of quadrants: do not collect that point more than once

Things that I know won't work:

• nearest point along just one axis, say x, can be very far away (circled green in image)
• second closest point in a quadrant (circled purple in image) may be closer than points in the other quadrants. A simple search for the red point's 4 nearest neighbors will collect this point I don't want.

EDIT: version of accepted answer's code, and timings

I have a list of coordinates a. I want to specify a radius r and then list all points on a 2D grid within the specified radius of any point in a, together with the minimum distance of each of those grid-points to any point in a. Since a is substantial (~1000-2000 points) I want to be as efficient as possible.

My implementation so far uses this code to list all coordinates within the given radius of a point; then iterates that over all coordinates in a; then flattens the output, takes a set (there will be MANY duplicates) - this is the set of coordinates within the specified radius of any point on the line - then calculates the minimum distances of that set to a using scipy.spatial.distance.cdist:

I have two numpy arrays both having coordinates of some points. One array has the coordinates of a cutting planar surface (cutting_surf) and another one has coordinates of some points (points):

Given a dataframe df as follows:

I have a class component that calculates the Euclidean distance between the last elements in arrays within 2 dictionaries. One dictionary contents the tracking trajectory of blobs (r), and the other have the updated values of the blobs(b). The methods of the class find the appeared or disappeared trajectories based on Euclidean distance. Finally, they reorder the r dictionary on their best match with the b dictionary.

I tested the functionality in this colab notebook and it works as expected, but when I implement it on my code the program gets slow.

1. Is there I way I can improve the speed of this class?
2. Is there a better approach to solve this problem? What is it?

Thank you.

i have a dataframe which consists of 472 rows and 32 columns and it looks like this:

Can someone help me please on how to generate a weighted adjacency matrix from a numpy array based on euclidean distance between all rows, i.e 0 and 1, 0 and 2,.. 1 and 2,...?

Given the following example with an input matrix(5, 4):