From Nikolaus.Hansen at lri.fr Mon May 22 16:55:15 2017
From: Nikolaus.Hansen at lri.fr (Nikolaus Hansen)
Date: Mon, 22 May 2017 16:55:15 +0200
Subject: [Bbob-discuss] Understanding the Output-Files
In-Reply-To: <8d25d04c-bfc9-c203-2503-3f9571077056@googlemail.com>
References: <8d25d04c-bfc9-c203-2503-3f9571077056@googlemail.com>
Message-ID:
Hi David,
The entry
2:295|-8.1e-09
is a very short summary of a single run/trial: it refers to a run made on instance 2 which took overall 295 evaluations and reached a best precision/distance to the final target f-value of -9.1e-09, i.e. it hit and went below the final target.
> On 3 Mar 2017, at 03:27, David Salewski wrote:
>
> Dear BBOBies,
>
> I was doing some test benchmarks for an algorithm and I tried to evaluate the results. I looked into the 'bbobexp_fX.info' files, but I can't seem to understand the meaning of some numbers.
>
> data_f1\bbobexp_f1_DIM2.dat, 1:20|3.5e-01, 2:20|5.2e-01, 3:20|1.4e-01, 4:20|3.3e+00, 5:20|2.9e+00, 41:20|3.2e+00, 42:20|3.3e+00, 43:20|2.1e+00, 44:20|1.8e+00, 45:20|2.1e+00, 46:20|2.5e-01, 47:20|1.4e+00, 48:20|1.0e+00, 49:20|2.7e-01, 50:20|8.8e-01
>
> data_f1/bbobexp_f1_DIM2.dat, 1:283|-9.7e-09, 2:295|-8.1e-09, 3:295|-8.5e-09, 4:211|-9.0e-09, 5:223|-9.5e-09, 41:247|-6.1e-09, 42:283|-6.6e-09, 43:199|-2.9e-09, 44:295|-1.7e-09, 45:283|-7.8e-09, 46:265|-4.4e-09, 47:229|-6.1e-09, 48:253|-7.3e-09, 49:253|-7.2e-09, 50:313|-6.6e-09
>
> data_f1/bbobexp_f1_DIM2.dat, 1:10|5.2e+000, 2:10|3.9e+001, 3:10|2.0e+001, 4:10|2.4e-001, 5:10|6.7e+000, 41:10|1.8e+001, 42:10|5.4e+000, 43:10|1.9e+001, 44:10|2.5e+001, 45:10|9.7e+000, 46:10|5.2e+000, 47:10|4.3e-001, 48:10|6.9e-001, 49:10|2.4e+000, 50:10|1.3e+000
>
> These three examples are from three different algorithms (the example ARS, CMA-CSA from last year and the one I'm working with). If we take "1:20|3.5e-01" for example, what does each number mean? I found out that the first number is the number of the instance, which is running, but I can't seem to figure out the other two ('20' and '3.5e-01'). Especially interesting would be why the '20' (or the '10' for the other algorithm) stays the same for all instances, but in the CMA-example, the number changes for the most part.
I assume the algorithms define the maximal number of evaluations to be 10 or 5 times dimension, respectively.
Hope that helps,
Niko
-------------- next part --------------
An HTML attachment was scrubbed...
URL: