Using swap_method = 'bandit'.
With a dataset smaller than the default batch size, we consistently run into out of bound errors.
And with datasets larger than the default batch size, only the first hundred elements of the full array are indexed, instead of correctly indexing on the batch itself.
Using swap_method = 'bandit'. With a dataset smaller than the default batch size, we consistently run into out of bound errors. And with datasets larger than the default batch size, only the first hundred elements of the full array are indexed, instead of correctly indexing on the batch itself.
`File ~/miniconda3/envs/mlplot-dev/lib/python3.9/site-packages/gam/clustering.py:842, in KMedoids._swap_pairs(self, X, d, a_swap, dist_func, idx_ref, n_used_ref, mu_x, sigma_x, D, E, Tih_min, h_i) 839 K_jih[idx] = np.minimum(diff_jh[idx], 0) 841 idx = np.where(diff_ji == 0) --> 842 K_jih[idx] = np.minimum(d_jh[idx], E[idx]) - D[idx] 844 # base-line update of mu and sigma 845 mu_x[h, i] = ((n_used_ref * mu_x[h, i]) + np.sum(K_jih)) / ( 846 n_used_ref + self.batchsize 847 )
IndexError: index 51 is out of bounds for axis 0 with size 50`