• PNAS Physics Portal
  • Science Sessions: The PNAS Podcast Program

Robust continuous clustering

  1. Vladlen Koltunb
  1. aDepartment of Electrical and Computer Engineering, University of Maryland, College Park, MD 20740;
  2. bIntel Labs, Santa Clara, CA 95054
  1. Edited by David L. Donoho, Stanford University, Stanford, CA, and approved August 7, 2017 (received for review January 13, 2017)

  1. Fig. 2.

    Runtime comparison of RCC-DR with AP and LDMGI. Runtime is evaluated as a function of dataset size, using randomly sampled subsets of different sizes from the MNIST dataset.

  2. Fig. 3.

    Visualization of RCC output on the MNIST dataset. (A) Ten randomly sampled instances <mml:math><mml:msub><mml:mi>??</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math>??i from each large cluster discovered by RCC, one cluster per row. (B) Corresponding representatives <mml:math><mml:msub><mml:mi>??</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math>??i from the learned representation <mml:math><mml:mi>??</mml:mi></mml:math>??. (C) Two random samples from each of the small outlying clusters discovered by RCC.

  3. Fig. S3.

    Visualization of RCC output on the Coil-100 dataset. (A) Ten randomly sampled instances <mml:math><mml:msub><mml:mi>??</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math>??i from each of 10 clusters randomly sampled from clusters discovered by RCC, one cluster per row. (B) Corresponding representatives <mml:math><mml:msub><mml:mi>??</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math>??i from the learned representation <mml:math><mml:mi>??</mml:mi></mml:math>??.

  4. Fig. 4.

    (A–C) Visualization of the representations learned by RCC (A) and the best-performing prior algorithms, LDMGI (B) and N-Cuts (C). The algorithms are run on 5,000 randomly sampled instances from the MNIST dataset. The learned representations are visualized using t-SNE.

  5. Fig. S1.

    (A and B) Robustness to hyperparameter settings on the YaleB (A) and Reuters (B) datasets.

  6. Fig. S2.

    Robustness to dataset imbalance.

Online Impact

                                      1. 99132880 2018-01-23
                                      2. 802899879 2018-01-23
                                      3. 295573878 2018-01-23
                                      4. 352668877 2018-01-23
                                      5. 984633876 2018-01-23
                                      6. 545928875 2018-01-23
                                      7. 976569874 2018-01-23
                                      8. 871324873 2018-01-23
                                      9. 263462872 2018-01-23
                                      10. 577161871 2018-01-23
                                      11. 255603870 2018-01-23
                                      12. 117346869 2018-01-23
                                      13. 90982868 2018-01-23
                                      14. 663415867 2018-01-23
                                      15. 793874866 2018-01-23
                                      16. 843582865 2018-01-23
                                      17. 864971864 2018-01-22
                                      18. 258841863 2018-01-22
                                      19. 957295862 2018-01-22
                                      20. 553518861 2018-01-22