Issues

• BanditPam installation failed

  ThrunGroup/BanditPAM-python#99 · Filter by author pavankamath99 opened on Dec 3, 2021
  1

• Add random seed to make_sigma_plots.py for reproducibility

  ThrunGroup/BanditPAM-python#93 · Filter by author motiwari opened on Jun 11, 2020

• Need to return medoids in unshuffled dataset

  ThrunGroup/BanditPAM-python#91 · Filter by author motiwari opened on Jun 11, 2020

• Some sort of rare deadlock

  ThrunGroup/BanditPAM-python#89 · Filter by author motiwari opened on Jun 4, 2020

• Need better error handling for threads -- many fail silently

  ThrunGroup/BanditPAM-python#88 · Filter by author motiwari opened on Jun 3, 2020

• Idea: Make \sigma_i adaptive based on \mu_i

  ThrunGroup/BanditPAM-python#77 · Filter by author motiwari opened on May 25, 2020

• Theoretical Intuition: Need k to be larger than the number of "natural" clusters

  ThrunGroup/BanditPAM-python#76 · Filter by author motiwari opened on May 24, 2020

• Bound on swap steps and its relation to NPHardness

  ThrunGroup/BanditPAM-python#69 · Filter by author motiwari opened on May 23, 2020

• Possible explanation of why CSH doesn't work well

  ThrunGroup/BanditPAM-python#68 · Filter by author motiwari opened on May 23, 2020

• Create an Experiment() class

  ThrunGroup/BanditPAM-python#61 · Filter by author motiwari opened on May 22, 2020

• Use argmin instead of np.where everywhere

  ThrunGroup/BanditPAM-python#56 · Filter by author motiwari opened on May 12, 2020

• UCB Swap step still depends on k

  ThrunGroup/BanditPAM-python#55 · Filter by author motiwari opened on May 12, 2020

• Note: L1 distance causes \sigma to become huge in scRNA

  ThrunGroup/BanditPAM-python#54 · Filter by author motiwari opened on May 11, 2020

• Might be able to optimize loop over swaps even more in cost_difference_fn...

  ThrunGroup/BanditPAM-python#48 · Filter by author motiwari opened on Apr 15, 2020

• Win/Note: Empirically, the number of points for which we compute exactly does not scale with p

  ThrunGroup/BanditPAM-python#43 · Filter by author motiwari opened on Apr 14, 2020

• Might want 2 different random seeds

  ThrunGroup/BanditPAM-python#38 · Filter by author motiwari opened on Apr 13, 2020
  1

• Gotcha: Right now can only perform 1 swap at each step

  ThrunGroup/BanditPAM-python#37 · Filter by author motiwari opened on Apr 13, 2020

• Gotcha: Outliers may introduce significant noise / variance in sampling

  ThrunGroup/BanditPAM-python#36 · Filter by author motiwari opened on Apr 13, 2020

• Bug: Optimization: candidate condition

  ThrunGroup/BanditPAM-python#35 · Filter by author motiwari opened on Apr 11, 2020

• Idea: Use better single best-arm identification approaches

  ThrunGroup/BanditPAM-python#3 · Filter by author motiwari opened on Mar 30, 2020

• Idea: Can we adaptively sample reference points?

  ThrunGroup/BanditPAM-python#2 · Filter by author motiwari opened on Mar 30, 2020

• Idea: how much can we change p?

  ThrunGroup/BanditPAM-python#1 · Filter by author motiwari opened on Mar 30, 2020