# Revision history [back]

### facenet math

Hello, i hope this is not too off topic but i want to try anyway please.

Pre conditions I have: * database filled with 128 floating vectors for different persons * the vector is a unit vector (length of 1)

So i am able to calculate the distance for a given vector on the database (apache solr) successfully. My naive approach calculates all distances and sorts by distance - which is working correctly.

Now my question: Is it possible to limit the records for a given max distance without calculating the distance for all records? I want to use fixed fields(no db functions) to make this prediction.

I tried so far : mean, variance, using anchor points but i see no correlation with the calculated distance. Sorry - i am still learning but math is always a bit difficult for me (i try to improve).

Thank you + Greetings, Holger

### facenet math

Hello, i hope this is not too off topic but i want to try anyway please.

Pre conditions I have: * have:

• database filled with 128 floating vectors for different persons * persons
• the vector is a unit vector (length of 1)

So i am able to calculate the distance for a given vector on the database (apache solr) successfully. My naive approach calculates all distances and sorts by distance - which is working correctly.

Now my question: Is it possible to limit the records for a given max distance without calculating the distance for all records? I want to use fixed fields(no db functions) to make this prediction.

I tried so far : mean, variance, using anchor points but i see no correlation with the calculated distance. Sorry - i am still learning but math is always a bit difficult for me (i try to improve).

Thank you + Greetings, Holger

### facenet math

Hello, i hope this is not too off topic but i want to try anyway please.

Pre conditions I have:

• database filled with 128 floating vectors for different persons
• the vector is a unit vector (length of 1)

So i am able to calculate the distance for a given vector on the database (apache solr) successfully. My naive approach calculates all distances for a request and sorts by distance - which is working correctly.

Now my question: Is it possible to limit the records for a given max distance without calculating the distance for all records? I want to use fixed fields(no db functions) to make this prediction.

I tried so far : mean, variance, using anchor points but i see no correlation with the calculated distance. Sorry - i am still learning but math is always a bit difficult for me (i try to improve).

Thank you + Greetings, Holger

### facenet math

Hello, i hope this is not too off topic but i want to try anyway please.

Pre conditions I have:

• database filled with 128 floating vectors for different persons
• the vector is a unit vector (length of 1)

So i am able to calculate the distance for a given vector on the database (apache solr) successfully. My naive approach calculates all distances for a request and sorts by distance - which is working correctly.

Now my question: Is it possible to limit the records the db needs to scan for a given max distance without calculating the distance for all records? I want to use fixed fields(no db functions) to make this prediction.

I tried so far : mean, variance, using anchor points but i see no correlation with the calculated distance. Sorry - i am still learning but math is always a bit difficult for me (i try to improve).

Thank you + Greetings, Holger