Jarnik's MST with query indexes kept sorted more cache-friendly?

gagolews / genieclust

Genie: Fast and Robust Hierarchical Clustering with Noise Point Detection - in Python and R

Other

58 stars 11 forks source link

before:

6
d=10, n=50000, iter=0
   random_state     timestamp     dataset      n   d              method  n_clusters n_threads  elapsed_time       ar        r       fm      afm       mi       nmi       ami     nacc      psi
0             1  1.594542e+09  g2mg_10_20  50000  10  Genie_0.3_nomlpack           2         6      3.901408  0.99904  0.99952  0.99952  0.99904  0.69091  0.996773  0.996773  0.99952  0.99944
d=100, n=50000, iter=0
   random_state     timestamp      dataset      n    d              method  n_clusters n_threads  elapsed_time            ar        r       fm           afm        mi       nmi       ami    nacc      psi
0             1  1.594542e+09  g2mg_100_20  50000  100  Genie_0.3_nomlpack           2         6     26.401537  3.264181e-07  0.49999  0.70693  3.938901e-07  0.000166  0.000479  0.000448  0.0006  0.00024
d=10, n=50000, iter=1
   random_state     timestamp     dataset      n   d              method  n_clusters n_threads  elapsed_time      ar       r      fm     afm        mi       nmi       ami    nacc     psi
0             2  1.594542e+09  g2mg_10_20  50000  10  Genie_0.3_nomlpack           2         6      4.014916  0.9992  0.9996  0.9996  0.9992  0.691244  0.997254  0.997254  0.9996  0.9996
d=100, n=50000, iter=1
   random_state     timestamp      dataset      n    d              method  n_clusters n_threads  elapsed_time            ar        r        fm           afm        mi       nmi       ami     nacc      psi
0             2  1.594542e+09  g2mg_100_20  50000  100  Genie_0.3_nomlpack           2         6     26.566503  4.832190e-07  0.49999  0.706845  5.830038e-07  0.000172  0.000495  0.000465  0.00072  0.00032
d=10, n=50000, iter=2
   random_state     timestamp     dataset      n   d              method  n_clusters n_threads  elapsed_time       ar        r       fm      afm        mi       nmi       ami     nacc      psi
0             3  1.594542e+09  g2mg_10_20  50000  10  Genie_0.3_nomlpack           2         6      3.968966  0.99912  0.99956  0.99956  0.99912  0.691075  0.997011  0.997011  0.99956  0.99952
d=100, n=50000, iter=2
   random_state     timestamp      dataset      n    d              method  n_clusters n_threads  elapsed_time            ar        r       fm           afm        mi       nmi       ami    nacc      psi
0             3  1.594542e+09  g2mg_100_20  50000  100  Genie_0.3_nomlpack           2         6     26.292954  5.984283e-07  0.49999  0.70686  7.220242e-07  0.000236  0.000677  0.000648  0.0008  0.00034

d=10, n=50000, iter=0 random_state timestamp dataset n d method n_clusters n_threads elapsed_time ar r fm afm mi nmi ami nacc psi 0 1 1.594543e+09 g2mg_10_20 50000 10 Genie_0.3_nomlpack 2 6 3.3671 0.99904 0.99952 0.99952 0.99904 0.69091 0.996773 0.996773 0.99952 0.99944 d=100, n=50000, iter=0 random_state timestamp dataset n d method n_clusters n_threads elapsed_time ar r fm afm mi nmi ami nacc psi 0 1 1.594543e+09 g2mg_100_20 50000 100 Genie_0.3_nomlpack 2 6 24.093857 3.264181e-07 0.49999 0.70693 3.938901e-07 0.000166 0.000479 0.000448 0.0006 0.00024 d=10, n=50000, iter=1 random_state timestamp dataset n d method n_clusters n_threads elapsed_time ar r fm afm mi nmi ami nacc psi 0 2 1.594543e+09 g2mg_10_20 50000 10 Genie_0.3_nomlpack 2 6 3.933692 0.9992 0.9996 0.9996 0.9992 0.691244 0.997254 0.997254 0.9996 0.9996 d=100, n=50000, iter=1 random_state timestamp dataset n d method n_clusters n_threads elapsed_time ar r fm afm mi nmi ami nacc psi 0 2 1.594543e+09 g2mg_100_20 50000 100 Genie_0.3_nomlpack 2 6 26.052527 4.832190e-07 0.49999 0.706845 5.830038e-07 0.000172 0.000495 0.000465 0.00072 0.00032 d=10, n=50000, iter=2 random_state timestamp dataset n d method n_clusters n_threads elapsed_time ar r fm afm mi nmi ami nacc psi 0 3 1.594543e+09 g2mg_10_20 50000 10 Genie_0.3_nomlpack 2 6 3.782147 0.99912 0.99956 0.99956 0.99912 0.691075 0.997011 0.997011 0.99956 0.99952 d=100, n=50000, iter=2 random_state timestamp dataset n d method n_clusters n_threads elapsed_time ar r fm afm mi nmi ami nacc psi 0 3 1.594543e+09 g2mg_100_20 50000 100 Genie_0.3_nomlpack 2 6 26.088192 5.984283e-07 0.49999 0.70686 7.220242e-07 0.000236 0.000677 0.000648 0.0008 0.00034

gagolews / genieclust

Jarnik's MST with query indexes kept sorted more cache-friendly? #62