v0dro

Planning

v0dro

Ruby Association Grant 2015 Final Report

Author: Sameer Deshmukh

Email : [email protected]

GitHub: @v0dro

Contents

• Overview

v0dro

a=Daru::DataFrame.from_csv('spec/fixtures/sales-funnel.csv',headers: true)
=> 
#<Daru::DataFrame:84880950 @name = 42706995-535d-4181-9a13-9ba2f4f65b3f @size = 17>
              account    manager       name      price    product   quantity        rep     status 
         0     714466 Debra Henl Trantow-Ba      30000        CPU          1 Craig Book  presented 
         1     714466 Debra Henl Trantow-Ba      10000   Software          1 Craig Book  presented 
         2     714466 Debra Henl Trantow-Ba       5000 Maintenanc          2 Craig Book    pending 
         3     737550 Debra Henl Fritsch, R      35000        CPU          1 Craig Book   declined 
         4     146832 Debra Henl Kiehn-Spin      65000        CPU          2 Daniel Hil        won 
         5     218895 Debra Henl  Kulas Inc      40000        CPU          2 Daniel Hil    pending 

v0dro

require 'daru'
require 'gnuplotrb'
require 'statsample-timeseries'

index = Daru::DateTimeIndex.date_range :start => '2012', :periods => 500
iteration_count = 500
actual_values = Daru::Vector.new(iteration_count.times.map { |j|  -0.37727 + j * j * 0.00001}, index: index)
noisy_data    = Daru::Vector.new(actual_values.map { |actual_val| rand * 0.6 - 0.3 + actual_val }, index: index)

GnuplotRB::Plot.new(

v0dro

Finished in 48.920916s, 8.2378 runs/s, 25.8376 assertions/s.

  1) Failure:
StatsampleReliabilityIccTestCase#test_: Statsample::Reliability::ICC with Shrout & Fleiss denominations, should icc(2,1) confidence interval should be correct.  [/home/sameer/gitrepos/statsample/test/test_reliability_icc.rb:81]:
Expected |0.019 - -0.02227501611508347| (0.04127501611508347) to be <= 0.001.


  2) Failure:
StatsampleReliabilityIccTestCase#test_: Statsample::Reliability::ICC McGraw and Wong ICC Type icc_a_k should bounds be equal.  [/home/sameer/gitrepos/statsample/test/test_reliability_icc.rb:186]:
Expected |0.98156527598367 - 0.9759274278050467| (0.005637848178623339) to be <= 0.001.

v0dro

  1) Error:
StatsampleFactorTestCase#test_covariance_matrix:
NoMethodError: private method `[]=' called for #<Matrix:0xa1667a0>
    /home/sameer/gitrepos/statsample/lib/statsample/factor/pca.rb:156:in `block (2 levels) in component_matrix_covariance'
    /home/sameer/gitrepos/statsample/lib/statsample/factor/pca.rb:155:in `times'
    /home/sameer/gitrepos/statsample/lib/statsample/factor/pca.rb:155:in `block in component_matrix_covariance'
    /home/sameer/gitrepos/statsample/lib/statsample/factor/pca.rb:154:in `times'
    /home/sameer/gitrepos/statsample/lib/statsample/factor/pca.rb:154:in `component_matrix_covariance'
    /home/sameer/gitrepos/statsample/lib/statsample/factor/pca.rb:145:in `component_matrix'
    /home/sameer/gitrepos/statsample/test/test_factor.rb:25:in `test_covariance_matrix'

v0dro

NMatrix: an efficient storage and linear algebra library for Ruby
 HB input and fortran format parser:
  Made a new class for parsing matrices stored in the Harwell Boeing file format 
  and loading the data as an nmatrix. Also wrote a separate class for parsing FORTRAN
  number formats, which tells the data type and storage of numbers in a text file.
  
  Link:
  https://github.com/SciRuby/nmatrix/commit/b4d3c25aa8d32f9bd20b6c0ac4b10b3f6933f2c4

 LAPACK free matrix inverse, #round, Rational division fix:

v0dro

You first have a function say `__solve__` being called from ruby code.

You then declare the equivalent C function that will be called from the C code that will be executed when `__solve__` is called.
These declarations (which look like `rb_define_function_blah_blah`) are kept inside a function called `void Init_nmatrix()` in the file `ruby_nmatrix.c`.

Say the C function is called `nm_solve`, then the declaration looks like `rb_define_private_method(cNMatrix, "__solve__", (METHOD)nm_solve, 4);`.

This `nm_solve` function should be declared in the file `ruby_nmatrix.c`.

Then you have a function called `nm_solve` in math.cpp whose sole function is to take shit (arguments) from `__solve__`, make sense of them (if any) and then call a C function called `nm_math_solve` whose sole job is to do the `ttable` stuff, and with those helpful macros, generate a C++ function call of the appropriate data type.

v0dro

# Note : The GLM results from R match those from statsample but do not match those of statsample-glm. Here is my R code.

> csv = read.csv("/home/sameer/github_repos/statsample/test/fixtures/test_binomial.csv")

> attach(csv)
 
> glm(y ~ +a+b+c, family = binomial)

# => Output
Call:  glm(formula = y ~ +a + b + c, family = binomial)

v0dro

'apt-get' install atlas and its dev files (I installed libatlas3-base, libatlas-base-dev, libatlas3gf-base, and libatlas-dev iirc. Not sure if all of these are needed; I just grabbed everything that looked relevant.).

Followed by: export CPLUS_INCLUDE_PATH=$CPLUS_INCLUDE_PATH:/usr/include/atlas before compiling nmatrix from source. (If I leave out the export, I have the same error shown in the gist linked in the OP.)

So at least for me, it was just a matter of ubuntu putting the headers in a different location.