> I am building a system that will be able to compare electrical brushes
> (graphite electrical brush).
> Basically the comparison will give the level of similarity and also
> highlight the differences between two brushes.
>
> The main problem is the quantity of brushes; i am talking about 500000
> brushes. So, i have to make 125 billions of comparisons (combinatory
> analysis C500000,2) in order to know the level of similarity among
> brushes.

I don't think I agree with your approach/reasoning.

For example to find all brushes which are identical to each other (in 1
dimension), it would be very easy get the list of distinct values and
then build a result which list the brushes which match or lists
resulting brush-pairs.

Could you please explain the types of analysis which would be required.

> * I have to store a reference for the two brushes and also the level

of

> similarity. I create a table like this
>
> |brush1(integer) | brush2(integer) | similarity (smallint) |
>

|-----------------------------------------------------------------------
--

> ----------|
> | 500000 | 34999 | 98
> |
> | 365 | 34999 | 99
> |
> | 1000 | 349 | 56
> |
> | 1000 | 365 | 75
> |

Is similarity determined/calculated using one or multiple factors?

Can you explain how similarity is determined/calculated?


Sean