C1_W4_Assignment - Choosing number of planes

Course Q&A Natural Language Processing NLP with Classification and Vector Spaces
1

I am not able to understand the logic given for choosing no. of planes in the assignment which is

Each plane divides the space to 2 parts.
So n planes divide the space into 2^{n} hash buckets.

Also choosing 16 vectors/bucket is arbitary ? why not go with 20 vectors/bucket or 32 vectors/bucket?

2

1-plane-> 2 spaces
2-planes → 4 spaces
…

The more planes you create the more computationally expensive it becomes to run, so you need something not very large but which is still able to divide the space into meaningful separate spaces.

2 Likes