Tutorial¶

Load some test data

In [1]:

import numpy as np

import numpy as np

In [2]:

labeled_grid = np.load("../data/test.npy")

labeled_grid = np.load("../data/test.npy")

Labels:

• -1 corresponds to background
• values > 0 are unique labels (that can be named later on)

In [12]:

np.unique(labeled_grid)

np.unique(labeled_grid)

Out[12]:

array([-1,  0,  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], dtype=int8)

Analysis¶

Initialize analysis

In [4]:

from subdomain import SubDomain

# generate labels for the 'categories'
max_label = labeled_grid.max()
labels = [f"Label{i + 1}" for i in range(max_label + 1)]

my_analysis = SubDomain(labeled_grid, labels=labels)

from subdomain import SubDomain # generate labels for the 'categories' max_label = labeled_grid.max() labels = [f"Label{i + 1}" for i in range(max_label + 1)] my_analysis = SubDomain(labeled_grid, labels=labels)

Calculate the neighborhoods by

• by binning the data
• calculating the neighborhood composition by aggregating the neighbors per bin

Multiple aggregation exist; a box kernel (square neighborhood), a circular neighborhood, or a Gaussian-weighted neighborhood.

In [5]:

my_analysis.calculate_neighborhoods(20, 4, neighborhood="gaussian", sigma=2)

my_analysis.calculate_neighborhoods(20, 4, neighborhood="gaussian", sigma=2)

Next, we cluster the neighborhoods using either

• k-means
• Gaussian Mixture Models (GMM)

In [6]:

my_analysis.cluster_neighborhoods(10, "gmm")

my_analysis.cluster_neighborhoods(10, "gmm")

We can also calculate the neighborhood composition per domain using [subdomain.SubDomain.identify_domains][]

In [7]:

my_analysis.domain_neighborhoods()

my_analysis.domain_neighborhoods()

Out[7]:

label	Label1	Label2	Label3	Label4	Label5	Label6	Label7	Label8	Label9	Label10	...	Label17	Label18	Label19	Label20	Label21	Label22	Label23	Label24	Label25	Label26
domain
0	0.001624	0.065288	0.080585	0.032000	0.081279	0.018805	0.048605	0.033070	0.016263	0.129040	...	0.026330	0.047621	0.026373	0.014849	0.046480	0.024611	0.038204	0.018816	0.041110	0.041122
1	0.012657	0.052411	0.072298	0.021331	0.077497	0.018372	0.033779	0.026296	0.052749	0.095195	...	0.023371	0.033507	0.028430	0.038381	0.045500	0.028812	0.021597	0.037075	0.053157	0.055448
2	0.003359	0.068107	0.074232	0.023434	0.044248	0.021461	0.036332	0.061972	0.017276	0.114444	...	0.019528	0.033728	0.024502	0.010430	0.058309	0.025296	0.054180	0.019729	0.064514	0.018602
3	0.011150	0.057753	0.080759	0.053148	0.076515	0.019053	0.062603	0.023532	0.024282	0.063858	...	0.036084	0.053840	0.029923	0.016450	0.051570	0.022279	0.030938	0.022119	0.047639	0.036553
4	0.003673	0.057070	0.080949	0.042781	0.065570	0.016284	0.041995	0.030810	0.014708	0.073270	...	0.066498	0.090890	0.025439	0.012175	0.042116	0.019180	0.035850	0.012677	0.051573	0.014765
5	0.002031	0.044589	0.076663	0.071872	0.103237	0.015885	0.090847	0.025130	0.012052	0.063133	...	0.039378	0.057717	0.026132	0.020005	0.043232	0.016410	0.030346	0.015004	0.043592	0.012658
6	0.026367	0.078345	0.061955	0.030577	0.063722	0.029547	0.040529	0.023593	0.041371	0.112597	...	0.022208	0.040046	0.025232	0.020854	0.040751	0.024288	0.033765	0.023404	0.032950	0.098800
7	0.002285	0.049217	0.061154	0.040488	0.066936	0.016912	0.060051	0.030216	0.012240	0.054422	...	0.050874	0.059475	0.040841	0.012291	0.045739	0.017270	0.019287	0.020274	0.035156	0.012764
8	0.009198	0.081101	0.077618	0.024898	0.045169	0.020570	0.033011	0.034285	0.021205	0.123622	...	0.020682	0.044443	0.024559	0.021055	0.049621	0.030381	0.077404	0.018744	0.060662	0.046249
9	0.000017	0.066684	0.108067	0.041562	0.127276	0.016493	0.065538	0.025038	0.000055	0.165997	...	0.012203	0.039764	0.020751	0.000033	0.066617	0.028745	0.028902	0.000087	0.016504	0.060991

10 rows × 26 columns

In [8]:

my_analysis.domain_composition()

my_analysis.domain_composition()

Out[8]:

label	0	1	2	3	4	5	6	7	8	9	...	16	17	18	19	20	21	22	23	24	25
domain
0	0.000038	0.078755	0.087857	0.029209	0.090168	0.016315	0.043005	0.028006	0.015916	0.167429	...	0.020060	0.036116	0.025010	0.017782	0.044951	0.025794	0.045310	0.019840	0.034631	0.036945
1	0.007109	0.042191	0.046925	0.013724	0.050728	0.030186	0.036552	0.036713	0.064393	0.122070	...	0.016302	0.012238	0.041783	0.066331	0.057793	0.041681	0.028758	0.027068	0.040311	0.097988
2	0.000580	0.097164	0.078964	0.016853	0.041356	0.015815	0.025490	0.059007	0.020175	0.162090	...	0.009075	0.022869	0.017619	0.010001	0.062022	0.018824	0.074410	0.018346	0.071263	0.012047
3	0.006150	0.086911	0.124547	0.058332	0.095323	0.012409	0.064157	0.013131	0.035982	0.066896	...	0.019841	0.044481	0.023131	0.020541	0.055055	0.020052	0.032204	0.021145	0.039973	0.044268
4	0.001115	0.078874	0.128774	0.036863	0.065853	0.009573	0.028191	0.018193	0.016997	0.078960	...	0.054498	0.152524	0.016005	0.013438	0.034362	0.013205	0.040201	0.009349	0.049832	0.008873
5	0.000251	0.051199	0.122248	0.096413	0.164041	0.007873	0.157941	0.010940	0.009520	0.052987	...	0.018777	0.042607	0.015093	0.021663	0.035918	0.009597	0.026665	0.010200	0.032172	0.006686
6	0.022404	0.105759	0.060104	0.023730	0.064653	0.021449	0.032687	0.012835	0.059045	0.132350	...	0.011065	0.025618	0.017940	0.025940	0.038136	0.021090	0.034141	0.022418	0.021181	0.166194
7	0.000435	0.065841	0.086445	0.033251	0.076353	0.010582	0.062482	0.017602	0.014012	0.051073	...	0.034808	0.054301	0.037613	0.014735	0.045473	0.013635	0.017377	0.018137	0.027684	0.007904
8	0.004490	0.117074	0.089714	0.017754	0.037720	0.013390	0.019398	0.019794	0.024475	0.165792	...	0.008867	0.033971	0.016987	0.028579	0.047329	0.025418	0.131023	0.015904	0.062249	0.041644
9	0.000000	0.051632	0.082217	0.034798	0.101599	0.021520	0.079668	0.024459	0.000000	0.216641	...	0.006927	0.046658	0.020616	0.000000	0.073440	0.010873	0.040615	0.000000	0.018684	0.107252

10 rows × 26 columns

Plotting¶

We can plot the neighborhood composition

In [9]:

fig = my_analysis.plot_neighborhood_heatmap()

fig = my_analysis.plot_neighborhood_heatmap()

... and the domains

In [10]:

fig = my_analysis.plot_domains()

fig = my_analysis.plot_domains()