From: Bill Kelly <billk@...>
Date: 2008-02-20T18:08:03+09:00
Subject: Re: The Smallest Circle (#157)

------=_NextPart_000_0117_01C8735C.CD2F6540
Content-Type: text/plain;
	format=flowed;
	charset="iso-8859-1";
	reply-type=original
Content-Transfer-Encoding: 7bit


From: "ThoML" <micathom@gmail.com>
> @bill
>> Would be interesting to calculate how much my crude hack tends to
>> deviate from the perfect solution(s).
> 
> You might be more interested in speed.

Hi Thomas,

I wasn't able to benchmark your solution because I couldn't
find the #combination method?

These are the times I get running the other solutions,
10 runs of 1000 points each.  Each run, every solution is
fed the exact same point data.

$ ruby 157_benchmark.rb
             user     system      total        real
FRANK   23.469000   0.297000  23.766000 ( 23.781000)
JUSTIN   6.765000   0.000000   6.765000 (  6.812000)
LIONEL   5.704000   0.000000   5.704000 (  5.703000)
DOUG     3.797000   0.031000   3.828000 (  3.829000)
PHILIPP  1.390000   0.016000   1.406000 (  1.422000)
BILL     0.281000   0.000000   0.281000 (  0.281000)

Looks like my speed is good... Still curious how my accuracy
compares though... :)

(Benchmark code is attached...)


Regards,

Bill


------=_NextPart_000_0117_01C8735C.CD2F6540
Content-Type: application/octet-stream;
	name="157_benchmark.rb"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
	filename="157_benchmark.rb"


require 'benchmark'

module PHILIPP ; eval(File.read('157_philipp_hofmann.rb'))  ;  end
module LIONEL ;  eval(File.read('157_lionel_bouton.rb'))    ;  end
module THOMAS ;  eval(File.read('157_thomas_ml.rb'))        ;  end
module DOUG ;    eval(File.read('157_doug_seifert.rb'))     ;  end
module FRANK ;   eval(File.read('157_frank_fischer.rb'))    ;  end
module JUSTIN ;  eval(File.read('157_justin_ethier.rb'))    ;  end
module BILL ;    eval(File.read('157_smallest_circle.rb'))  ;  end

namespaces = [
# THOMAS,   # points.combination ??
  FRANK,
  JUSTIN,
  LIONEL,
  DOUG,
  PHILIPP,
  BILL ]

# Some folks implemented generate_samples differently,
# but we want to run all solutions on the same dataset.
# We'll use an implementation equivalent to that
# provided by the quiz author: [rand, rand]

num_runs = 10
num_samples = 1000

def generate_samples(mod, coords)
  coords.map {|xy| mod::Point.new(*xy)}
end

# Pregenerate different point data for successive runs.
# (All solutions are fed the same point data per run.)

pointsets = {}
namespaces.each do |mod|
  pointsets[mod.name] = (1..num_runs).map do
    point_coords = (1..num_samples).map {[rand, rand]}
    generate_samples(mod, point_coords)
  end
end

solutions = {}
namespaces.each {|mod| solutions[mod.name] = Class.new { include mod }.new}

label_width = namespaces.map {|mod| mod.name.length}.max
Benchmark.bm(label_width) do |x|
  namespaces.each do |mod|
    name = mod.name
    solution = solutions[name]
    x.report(name) do
      num_runs.times do |i|
        points = pointsets[name][i]
        solution.encircle(points)
      end
    end
  end
end


------=_NextPart_000_0117_01C8735C.CD2F6540--