Ram-hours

[Heitor Murilo Gomes]

Good day,

I have a few assumptions about ram-hours, which I would be grateful if someone could confirm or refute.

1) If I am running moa without sizeofag.jar, then ram-hours will output negative values, which means ram-hours are not being measured.
2) Ram-hours are given in GBs, therefore if I want them on MBs all I have to do is multiply then by 1024.
3) Assuming 1 and 2 are true, then the following table would be correct. Columns 1 to 3 are original outputs, column 4 represents RAM-Hours (GB) * 1024.

learning evaluation instances	evaluation time (cpu seconds)	RAM-Hours (GB)	RAM-Hours (MB)	classifications correct (percent)
1000	229.357725	0.01	12.81	92.8
2000	510.4135304	0.05	50.21	86.5
3000	776.8969442	0.09	91.13	92.9
4000	1120.754443	0.15	157.28	95.2
5000	1458.588633	0.23	237.33	90.9
6000	1845.948178	0.33	333.29	90.2
7000	2259.964169	0.46	469.58	92
8000	2683.237874	0.61	625.60	97.6
9000	3090.976826	0.77	785.96	97.1
9324	3429.94257	0.90	922.67	96.1

4) Regarding experiment presented on the table above, it is reasonable to state: "model cost was 922.67 MB"

Best Regards,
Heitor

[Albert Bifet]

Hi Heitor,

1/ That's correct, if there is no sizeofag.jar present, the Ram-hours measures are negative

2/ Ram-hours are given in Gb/Hour

3/ Yes, if you have z Gb/Hour, you have 1024z Mb/Hour
4/ model cost was 922.67 MB/Hour

Cheers, Albert

--
You received this message because you are subscribed to the Google Groups "MOA users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moa-users+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Heitor Murilo Gomes]

Thank you,  Albert.

Best Regards,
Heitor

[Riham]

Hi Sir,

Please would you mind telling me if it is Gbytes/hour or Gbits/hour?

Kind Regards