There is a type of data consist of 2 parts: class id(integer) and subId(integer) ,
I want to encode each record of the data into a SDR with 2048 bits and 20 bits a bucket.
The bucket is devided into 2 parts:13bits represents the class id, and 7 bits represent the subId,. the class id is encoded by a categoryEncoder which means the class id is encoded with no overlap with other class id and the records in the same class with high semantic similarity .
With the 13bits representing the class id, 10 of 13 bits are used to encode the class id ,which means there will be about 286(13×2×11 =286 ) class ids can be represented.
With the 7bits representing the sub id, 5 of 7 bits will be used to encode the sub ids,which means there will be about 21( 7*6/2=21) sub ids can be represented.
Is the encoding method ok? any idea will be appreciated.