Pruning frequent/infrequent indices #13
Description
We should allow users to specify minimum and/or maximum frequencies of indices for various modes. When and index appears too often, or not often enough, drop it. This could also be done as a percentage (e.g., "drop the 5% most frequent items").
This comes at the cost of requiring additional traversals over the tensor data. Each time an index is pruned, it may make others frequent/infrequent, and thus we have to iterate until convergence.