MathGroup Archive 2006

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Speed challenge: Improve on integer frequencies from Count?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg67790] Re: Speed challenge: Improve on integer frequencies from Count?
  • From: dh <dh at metrohm.ch>
  • Date: Fri, 7 Jul 2006 07:13:40 -0400 (EDT)
  • References: <e8ir3t$s8o$1@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

Hi Gareth,
change the range to: {0, maxInt} so that we can use them as indices.Then 
setup a list of counters, initially set to 0. Use every random number as 
  index of the counter to increment.E.g.
maxInt = 100;
counters = Table[0, {maxInt}];
d = Table[Random[Integer, {1, maxInt}], {1000}];

(counters[[#]]++) & /@ d;

counters now contains the frequencies.
Daniel

Gareth Russell wrote:
> Hi,
> 
> A challenge for the efficiency gurus:
> 
> Given a list of integers whose possible values lie in the range 0 to 
> maxInt, with any given integer represented 0, 1 or multiple times, 
> e.g., what you would get from
> 
> Table[Random[Integer, {0, maxInt}], {1000}]
> 
> create a list of frequencies of values including ALL possible values 
> (i.e., not the basic output of Frequencies[], because that only 
> includes values present in the list.
> 
> The fastest I have been able to come up with is the obvious use of Count[]...
> 
> In[256]:=
> myFrequencies=Compile[{{myList,_Integer,1},{
>     maxInt,_Integer}},Table[Count[myList,i],{i,0,maxInt}]]
> 
> Out[256]=
> CompiledFunction[{myList,maxInt},Table[Count[myList,i],{i,0,maxInt}],-\
> CompiledCode-]
> 
> In[278]:=
> testList=Table[Random[Integer,{0,500}],{1000}];
> 
> In[281]:=
> Timing[Do[myFrequencies[testList,500];,{1000}]]
> 
> Out[281]=
> {2.3103 Second,Null}
> 
> However, it seems to me that this should be able to be improved upon, 
> because it scans the whole list maxInt times. One should be able to 
> scan the list once, and increment a frequency list in the right places, 
> perhaps with a For[] loop. But try as I might, I can't come up with a 
> version that is faster than the compiled Count[] version above.
> 
> Can anyone do it?
> 
> Gareth
> 


  • Prev by Date: Re: Speed challenge: Improve on integer frequencies from Count?
  • Next by Date: Re: Speed challenge: Improve on integer frequencies from Count?
  • Previous by thread: Re: Speed challenge: Improve on integer frequencies from Count?
  • Next by thread: Re: Speed challenge: Improve on integer frequencies from Count?