forked from M-Lin-DM/Uniform-Orientation-Sampling
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Uniform_Orientation_Sampling.py
60 lines (44 loc) · 2.04 KB
/
Uniform_Orientation_Sampling.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
import numpy as np
from tqdm import tqdm
from utils import normalize_rows, Pair_separation, Bubble_tea_potential, KNN_repulsion
from Plotting_tools import NDScatter
from Evaluations import KNNDistanceDistribution
class UniformOrientationSampling:
"""Main class in charge of specifying the parameters and running the optimization program.
Parameters
------------
dimensions: int
dimensionality of the embedding/ambient space. Dimensions above 20 haven't been tested.
pop_size: int
Number of points (ie orientations) to learn.
iterations: int
Number of passes over the set of pop_size points. Each iteration updates all positions once.
approach: str
algorithm to use for updating positions. 'KNN_repulsion' is the only one that has been shown to work reliably
"""
def __init__(self, dimensions=3, pop_size=20, iterations=1000, approach='KNN_repulsion'):
self.dimensions = dimensions
self.pop_size = pop_size
self.iterations = iterations
self.approach = approach
def initialize_emb(self):
dat = np.random.rand(self.pop_size, self.dimensions) - 0.5
return normalize_rows(dat)
def update_emb(self, V):
# calls a given updater, which typically cycles over all points, updating each once
if self.approach == "pair_separation":
updater = Pair_separation(0.05)
return updater.update_points(V)
elif self.approach == "bubble_tea_potential":
updater = Bubble_tea_potential(0.02)
return updater.update_points(V)
elif self.approach == "KNN_repulsion":
updater = KNN_repulsion(0.01, K=1)
return updater.update_points(V)
def run_optimizer(self):
embedding = self.initialize_emb()
# emb_list = []
for t in tqdm(range(self.iterations)):
embedding = self.update_emb(embedding)
# emb_list.append(embedding)
return embedding