Author: Robert Hyatt
Date: 11:06:28 09/09/02
Go up one level in this thread
On September 09, 2002 at 13:48:22, Miguel A. Ballicora wrote: >On September 09, 2002 at 00:18:28, Robert Hyatt wrote: > >>On September 08, 2002 at 20:46:44, Miguel A. Ballicora wrote: >> >>>On September 07, 2002 at 11:13:20, Robert Hyatt wrote: >>> >>>> >>>>Jose made a really good point about observed data vs measured data. After >>>>thinking about it for a bit, I decided that it is a point strong enough to >>>>change the way we think about "measured" and "observed". >>>> >>>>Some examples: >>>> >>>>speed. Impossible to measure. >>>> >>>>For example, your automobile (newer vehicles) compute speed by counting the >>>>revolutions of the tailshaft (output) of the transmission, then factoring in >>>>the rear-end ratio and the circumference of the rear wheels. It _computes_ >>>>the speed from that. >>>> >>>>A radar measures the frequency change in a radio signal as it bounces off >>>>a moving target and _computes_ the speed based on the frequency change. >>>> >>>>A GPS observes to "positions" in terms of lattitude and longitude, uses some >>>>geometry to compute the distance between them, and uses a clock to measure the >>>>time to cover that distance, and displays speed. >>>> >>>>So Speed can't be measured directly, it has to be computed. And this isn't a >>>>surprise since speed is defined as distance over time. >>>> >>>>Brightness. (of a light, not a person. :) ) >>>> >>>>This is a direct measure of an electrical signal produced by some sort of >>>>device (photo-resistor, photo-cell, optical transistor, etc) and then that >>>>voltage is used to compute a brightness level in Lumens... >>>> >>>>Loudness (sound). >>>> >>>>Ditto. >>>> >>>>NPS. >>>> >>>>nodes searched divided by time in seconds. Computed. >>>> >>>>Speedup >>>> >>>>one-processor time divided by the N-processor time. Computed >>>> >>>>We really don't have a lot of "observed" data nowadays. Some, yes. Where >>>>were you at 8pm last night. But more is computed... >>>> >>>>Which means if we start to define observed vs computed, we don't end up with >>>>very much in the "observed" column. >>>> >>>>In a chess program I can count nodes and "compute" time (end-time minus >>>>start-time) and then compute a nps value. I can measure run-time and compute >>>>speed-up. But I can't directly measure speed at all. >>> >>>Not really, NPS is a direct measure. You do not measure nodes, you "report" >>>them. >> >>Please tell me _exactly_ how to measure nps and report it. The only way I >>know of is to measure nodes, measure time, and divide the former by the latter. > >I did not say that I report nps, I report nodes. Nodes have no error. In error >propagation is treated as a constant because the error is +/- 0. > >>If NPS is "directly measured" then so is "speedup" since it requires two >>measured quantities and a division also. > >I do not know what speedup has to do with this discussion. I thought that we >moved on from the discussion about the paper and I hope that you are not trying >to justify the treatment of the data with the impossibility of "measuring" >certain parameters as direct data. no... the discussion turned to the impossibility of measuring certain things, such as speedup, NPS, and the like, since those have to be computed. But even nodes is not something that is easy to obtain in a SMP engine, for the reasons I have given already. And even in a non-SMP engine that doesn't print nodes with each PV change can cause problems in the test I used for the DTS paper... however, it struck me as interesting that much of the data we would call "observed" is not. Speed of an auto is one, because it has to be calculated. > > >> >> >>> In other words, there is no error in the measure of nodes. What you really >>>measure is the time to reach a certain number of "events" (in this case is >>>nodes). Later, you make a conversion = [constant]/[time]. After a conversion, >>>the measure does not become indirect because of the "calculation". You can >>>measure inches an later you express them in cm. Still a direct measure. All the >>>error depends on only one parameter measured. >> >> >> >>But velocity is not _measured_ directly. You measure time and distance >>and then are forced to compute velocity. Nps is _exactly_ the same. > >It is not the same Value = [x]/[y] >and Value = [constant]/[y] >When you propagate errors. >Nodes is different than distance in that respect. Only if you assume you can perfectly count them. In a SMP program this isn't a given. In a cluster it also isn't a given. So there _will_ be some error in the counting process, and it becomes just like speed. Even in time there is error on a computer because time is not accurate to the nearest nanosecond, so that even with perfect node counting, NPS will be but an estimate with an undefined error depending on the hardware and software platforms being used. > > >>In the context of measured or computed anyway. > >The fact that you have to compute something does not mean is indirect data. >If you really want to be phylosophically picky, there is no direct measure. Not >even distance as R. gibert pointed out, because you always compute something or >compare with a reference. In case of distance, it is always end-start even when >you use a ruler. > I agree totally... But it seems that in some cases we are "happy" and accept such measures using a ruler as raw data, even when the raw data might have been calculated from a different measurement (circumference vs diameter, etc.) >Regards, >Miguel > > > > >> >> >> >> >> >> >>> >>>Miguel >>> >>>> >>>>strange when you think about it... :)
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.