headers
Hervé Brönnimann | 1 Feb 02:23
Picon

Re: [Review] Review of the Accumulators library begins today Jan 29


On Jan 31, 2007, at 10:38 AM, Hans Meine wrote:
> On Wednesday, 31. January 2007 16:19, Ben FrantzDale wrote:
>>> Isn't this already a problem with sequences? Suppose you're  
>>> storing the
>>> current median, and a new value comes along. What's the new median -
>>> without checking all the previous values? There are statistical  
>>> estimates
>>> for this, but I don't know any exact way other than essentially  
>>> resorting
>>> the data and checking the new median.
>>
>> Good point. Combining partial results may not be possible for all
>> accumulated statistics or there may be a cost to making a statistic
>> "combinable". Perhaps a refinement of the Accumulator concept  
>> would be
>> needed.
>>
>> The idea of the Accumulator library is to implement online  
>> algorithms; is
>> there a corresponding word for online algorithms for which partial  
>> results
>> can be combined?
>
> Actually, I don't see any "online algorithms" whose partial results  
> cannot be
> combined.
>
> IIUC, what James pointed out is exactly that the median is not an  
> online
(Continue reading)

Permalink | Reply | 0 comments |
headers
Art Stamness | 1 Feb 02:22
Picon
Favicon

Re: Request for interest in new library : Bit Logic

(oops hit reply-to-sender, not list )

Begin forwarded message:

> From: Art Stamness <artstamness <at> yahoo.com>
> Date: January 31, 2007 5:11:21 PM PST
> To: me22 <me22.ca <at> gmail.com>
> Subject: Re: [boost] Request for interest in new library : Bit Logic
>
> On Jan 31, 2007, at 4:21 PM, me22 wrote:
>
>> On 1/31/07, Art Stamness <artstamness <at> yahoo.com> wrote:
>>> Any feedback, positive, negative, suggestions is always appreciated.
>>> Also if there is a better suggestion for a name, I am open to it. I
>>> realize the Boost.Logic library has a very different purpose than
>>> this one, and don't want them to be confused.
>>>
>> Boost.Bitwise, maybe?
>>
>> Typo: http://www.emulogic.com/bit_logic/bit_vector/index.html
>> "middle of the container is dones in"
>
> Thanks, corrected now.
>
>> Regarding the implementation, smallest_bit_storage would probably be
>> better off using Boost.Integer's facilities or a trick like is using
>> in uint_t to avoid requiring the specifically-sized types.
>
> Those types are limited by the largest integral type of the system.  
> This library is intended for modeling small and large types. I
(Continue reading)

Permalink | Reply | 2 comments |
headers
Hervé Brönnimann | 1 Feb 02:39
Picon

Re: [Review] Review of the Accumulators library begins today Jan 29

Matthias:  Forgive me, I haven't looked at your code yet.  I wrote a  
few accumulators, in my days, but I don't quite understand what you  
mean by "combine".  One may combine online algorithms, e.g. pass the  
result of a filter to a summation or average accumulator.  But I  
don't quite grasp what you might do by combining a median into  
another algorithm.  Median is not really an online algorithms:  It  
maintains some quantities (e.g. a sample), but only at certain times  
(periodically, or at the end) the result of the computation (median)  
is queried.   There is no continued output.  Or you could imagine  
feeding the median continuously to another accumulator (average  
median of stream?) but it stretches the imagination?

For sampling, which has been a focus of my research recently, there  
is of course the reservoir sampling, but a few other special-purpose  
deterministic sampling algorithms have been devised.  On my  
publication web page, (http://photon.poly.edu/~hbr/publis- 
themes.html, section 5 - shameless plug, sorry))  I give a few  
sampling algorithms.  I could program them, if I have the time.  The  
so-called Biased-L2 and DRS are much better at preserving  
correlations / association rules for transactional data sets.

I also meant to note in passing that it is well known in database  
community that sampling is bad for estimating sizes of join, and  
other quantities dependent on correlation.  There are sophisticated  
ways (not too hard to program, though) to estimate them:  the  
celebrated Alon, Gibbons, Mattias, and Szegedy trick of projecting  
onto a number of random vectors (if you're interested / inclined:   
The Space Complexity of Approximating the Frequency Moments. J.  
Comput. Syst. Sci. 58(1): 137-147 (1999), portal.acm.org/citation.cfm? 
id=237823).  They also fall within the area of data streams I
(Continue reading)

Permalink | Reply | 1 comment |
headers
Chris Weed | 1 Feb 03:31
Picon

Re: Request for interest in new library : Bit Logic

Hi,
I looked at the docs, but some of the links are broken on the first page.
I tried IE and Firefox.

I also found some of the examples confusing, such as clicking
bit_array -> select, and the example used a bit_vector.

Even more confusing is the lack of documentation of the classes'
template argument.

Also, If I can do the following
typedef bit_array<16> 16bits;
I would also expect to be able to get the size value at compile time
through a const, static value or enum, such as 16bits::size.
Chris

On 1/31/07, Art Stamness <artstamness <at> yahoo.com> wrote:
> I have been working on a library that is used for low level bit
> manipulation of integral types, and modeling of n-bit length types. I
> have written it as a template library, for inclusion with the Boost
> library ( hopefully).
>
> This is my public request for interest to see if anyone would be
> interested in such a library. This code runs under  Visual Studio
> and g++ ( linux, and Mac ) . I have posted a web page with
> documentation ( still a work in progress ), of most all the
> functionality in the library. There is also a tar ball there, that
> has been tested with Boost 1.33.1 .
>
> Documentation and source tar-ball can be found here :
(Continue reading)

Permalink | Reply | 1 comment |
headers
Cromwell Enage | 1 Feb 03:59
Picon
Favicon
Gravatar

Re: Request for interest in new library : Bit Logic

--- Art Stamness wrote:
> http://www.emulogic.com/bit_logic
> 
> Any feedback, positive, negative, suggestions is
> always appreciated.  

Can you illustrate the differences between your
library, std::bitset, and boost::dynamic_bitset?

                              Cromwell D. Enage

 
____________________________________________________________________________________
8:00? 8:25? 8:40? Find a flick in no time 
with the Yahoo! Search movie showtime shortcut.
http://tools.search.yahoo.com/shortcuts/#news
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Permalink | Reply | 1 comment |
headers
Roland Schwarz | 1 Feb 07:51
Picon

Re: request to patch libs/serialization/build/Jamfile.v2forsupport of msvc-6.5

Boris Gubenko wrote:
> I think (not sure!) you need a whitespace character before the trailing 
> semicolon.

Almost for sure!

Hint: Try bjam --debug-configuration
this should print out the found paths.

Roland

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Permalink | Reply | 1 comment |
headers
Matt Hurd | 1 Feb 07:55
Picon
Favicon

Re: [Review] Review of the Accumulators library begins today Jan 29

> On 01/02/07, james.jones <at> firstinvestors.com <james.jones <at> firstinvestors.com> wrote:
> From: Matthias Troyer <troyer <at> phys.ethz.ch>

<snip>

> Isn't this already a problem with sequences? Suppose you're storing the current median, and a new value
comes along. What's the new median - without checking all the previous values? There are statistical
estimates for this, but I don't know any exact way other than essentially resorting the data and checking
the new median.

You can get improvement to an on-line median by have storing some of
the central state.  If the distribution is kind and you get lucky you
can avoid a lot of checking of previous values giving nice
improvements.

Similar stuff works for min/max, quartiles and pecentiles.

For windowed esimates you need to account for values falling off and
those coming in but the same works.

Done wisely you can get an order of magnitude improvement for appropriate data.

$AUD 0.0002

Matt.
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Permalink | Reply | 0 comments |
headers
Roland Schwarz | 1 Feb 07:59
Picon

Re: boost_unit_test_framework cannot be built as DLL withmsvc-6.5

Gennadiy Rozental wrote:
> Boost.Test doesn't support DLLs for msvc 6.5. And have no plans for this to 
> change.

So what? Some of the test tests are explicitly making use of DLL's. 
Should these simply omitted for msvc-6.5? Shall I try to modify the Jamfile?

Then I think I almost racked down from where the error comes:

It seems that the compiler does not allow to embed a member
callback0<> m_callbackfn;
in an exported class. I assume that this callback is across DLL 
boundaries, yes? Perhaps someone with knowledge of the inner workings of 
Boost.Test can suggest a work around?

Roland

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Permalink | Reply | 15 comments |
headers
Roland Schwarz | 1 Feb 08:35
Picon

[thread] regressions for msvc-8.0 x64

I am unsure how this "regression" should be handled.

Since this _is_ a new compiler it is not really a regression,
but on the other hand the compilers toolset name is the same
as for x32.

1) Is there a way to explicitly markup as expected to fail
    conditioned on x64 ?

2) I do not have yet an x64 compiler so I will not be able
    to implement the tss cleanup for static linking for 1.34.
    I'll do this later.

Roland
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Permalink | Reply | 2 comments |
headers
Anthony Williams | 1 Feb 08:49
Picon
Favicon

Re: request to patch libs/serialization/build/Jamfile.v2forsupport of msvc-6.5

Roland Schwarz <roland.schwarz <at> chello.at> writes:

> Boris Gubenko wrote:
>> I think (not sure!) you need a whitespace character before the trailing 
>> semicolon.
>
> Almost for sure!

Thanks guys! I knew that, but didn't spot it. I'll do a run tonight now that's
fixed.

Anthony
--

-- 
Anthony Williams
Just Software Solutions Ltd - http://www.justsoftwaresolutions.co.uk
Registered in England, Company Number 5478976.
Registered Office: 15 Carrallack Mews, St Just, Cornwall, TR19 7UL

_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
Permalink | Reply | 0 comments |

Gmane