Similarity Metric

Here is a general idea for a universal memory cell implementation: if the current input is similar to the previous input minus the censors, then we have a variation previous input. I.e. if a new input is similar to previously encountered 0x1A2B4C5D6E7F and censors filter out symbols other than 0-9 and A-F, then we have a hexadecimal number.

The similarity metric should itself be affected by censors recursively, while censors themselves are created from bad input. I.e. if system classified 0x4ZZZ7 as a hexadecimal number and got slapped in the face, it should add similarity to 0x4ZZZ7 as a new censor.

I am not entirely sure what we are trying to achieve with this.