Re: [eclipse-users] Counting Backtracks

From: Joachim Schimpf (Independent Contractor) <jschimpf_at_cisco.com>
Date: Sun, 02 Mar 2008 20:20:01 +0000
Kish Shen wrote:
> kot23 wrote:
...
>> The output I have is like this:
>>
>>      : backtracks : 0
>> Found a solution with cost 335
>>      : backtracks : 901
>> Found a solution with cost 285
>>      : backtracks : 904
>> Found a solution with cost 270
>>      : backtracks : 915
>> Found a solution with cost 255
>> Found no solution with cost 240.0 .. 240.0
>>
>> So, my question is, what is the real number of backtracks, for finding
>> the last solution in the case above? Is it "915" backtracks or the sum of
>> each intermediate solution, ie: 0+901+904+914=2719 backtracks?
>>
>> Thanks,
>> Christoforidis Nikos
>>
>>   
> As stated in the documentation, the backtracks reported is the number of 
> backtracking steps used in the search routine, i.e. 915 is the total 
> number of backtracks performed by the search to find the last (optimal) 
> solution.

There is a slight pitfall here: what Kish says is only true as long
as you use the strategy 'continue' in bb_min (which is the default).

If you set the strategy option to 'restart' or 'dichotomic', then the
search predicate will be re-started after every new minimum found,
therefore it will restart with a backtrack counter of 0 every time,
and you will have to add the counts in order to get the total.


-- Joachim
Received on Sun Mar 02 2008 - 20:20:20 CET

This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET