lasso is not able to handle matrices instead of vectors

[ingenieroariel]

Here is an example that could be added to the test suite, for my problem I need to send matrices instead of vectors to the lasso solver:

using ParallelSparseRegression

dimensions, components, samples = 180, 100, 1500

dictionary = sprandn(dimensions, components, 0.1)
original_code = sprandn(components, samples, 0.1)
data = dictionary * original_code

# minimize \|data - dictionary*code\|_2^2 + \lambda\|code\|_1
# for variable code.
code = lasso(data, dictionary, 1)

# FIXME(Ariel): Is this the best way to check it found the real solution?
@assert(code = original_code)

[madeleineudell]

Generically, @assert(code = original_code) will NOT be true for a
solution to the lasso problem. The lasso is a biased estimator; the
recovered code will have smaller norm than the original code, even if it
has the same sparsity pattern. To test this solver, I would suggest
computing a solution to a particular instance using a known and trusted
solver (for example, cvx or tfocs in matlab, cvxpy or cvxopt in python) and
comparing the the solution returned by this package to that one. The
solutions should be close, up to the tolerance specified by the algorithm,
eg TOL = 1E-4; @assert(norm(recovered_code - cvxpy_recovered_code)<TOL)

On Fri, Feb 21, 2014 at 5:59 AM, Ariel Núñez notifications@github.comwrote:

Here is an example that could be added to the test suite, for my problem I
need to send matrices instead of vectors to the lasso solver:

using ParallelSparseRegression

dimensions, components, samples = 180, 100, 1500

dictionary = sprandn(dimensions, components, 0.1)
original_code = sprandn(components, samples, 0.1)
data = dictionary * original_code

minimize |data - dictionary*code|_2^2 + \lambda|code|_1

for variable code.

code = lasso(data, dictionary, 1)

FIXME(Ariel): Is this the best way to check it found the real solution?

@Assert(code = original_code)

Reply to this email directly or view it on GitHubhttps://github.com//issues/5
.

Madeleine Udell
PhD Candidate in Computational and Mathematical Engineering
Stanford University
www.stanford.edu/~udell

[ingenieroariel]

I used CVX in matlab and saved the matrices to a file:
http://ingenieroariel.com/static/admm.mat

The sizes of A, B_cvx and X are (100,100), (100,80) and (100,80):

using MAT
using ParallelSparseRegression

# Get sample data from a .mat file
file = matopen("admm.mat")
A = read(file, "A")
B_cvx = read(file, "B")
X = read(file, "X")
cost = read(file, "cost")
close(file)

# minimize \|data - dictionary*code\|_2^2 + \lambda\|code\|_1
# for variable code.
B = lasso(X, A, 1)

TOL = 1E-4; @assert(B == B_cvx)

I still get "ERROR: BoundsError()" when I call the lasso, most likely because it is using matrices instead of vectors, is there a good way to reformulate my problem before sending to the lasso? I remember reading something about kron product, will look it up.

[ingenieroariel]

I also checked that the following lasso admm implementation by Simon Lucey produced the same results in the file:

% Function to perform LASSO regression using Alternating Direction Method
% of Multipliers.
%
% arg min_{B} 0.5*||X - A*B||_{2}^{2} + gamma*||B||_{1}
%
% Usage:- [B,cost] = lasso_admm(X, A, gamma) 
%
% where:- <in>
%         b = bias vector
%         lambda = weighting on the l1 penalty
%         <out>
%         x = solution  
%
% Written by Simon Lucey 2012

function [B,cost] = lasso_admm(X, A, gamma)

% Get dimensions of B
c = size(X,2);
r = size(A,2); 

L = zeros(size(X)); % Initialize Lagragian to be nothing (seems to work well)
rho = 1e-4; % Set rho to be quite low to start with 
maxIter = 500; % Set the maximum number of iterations (make really big to ensure convergence)
I = speye(r); % Set the sparse identity matrix
maxRho = 5; % Set the maximum mu
C = randn(r,c); % Initialize C randomly

% Set the fast soft thresholding function
fast_sthresh = @(x,th) sign(x).*max(abs(x) - th,0);

% Set the norm functions
norm2 = @(x) x(:)'*x(:); 
norm1 = @(x) sum(abs(x(:))); 

cost = [];
for n = 1:maxIter
    F = (A'*A+rho*I);
    G = (A'*X + rho*C - L);
    % Solve sub-problem to solve B
    B = F \ G; 

    if n==115
        1 ==1
    end
    % Solve sub-problem to solve C
    C = fast_sthresh(B + L/rho, gamma/rho); 

    % Update the Lagrangian
    L = L + rho*(B - C);  

    %pause; 

    % Section 3.3 in Boyd's book describes strategies for adapting rho
    % main strategy should be to ensure that 
    rho = min(maxRho, rho*1.1); 

    % get the current cost
    cost(n) = 0.5*norm2(X - A*B) + gamma*norm1(B);

end

[madeleineudell]

This package is not designed to do regression on matrices. But if you want to solve minimize |AX - B|_2^2 + |X|_1, where X and B are matrices, you can reformulate the problem as a vector problem in the following way. Let b = vcat([B[:,i] for i in size(B,2)]) be the column vector that concatenates the columns of B, and let AA = vcat([A for i in size(B,2)]) be a matrix that vertically concatenates as many copies of A as there are columns in B. Then lasso(AA,b) gives a vector which vertically concatenates the columns of the matrix X you're looking for.

[ingenieroariel]

Okay, I have been able to test this module against an implementation in CVX and a simple lasso, it seems the result I get is a solution to the l2 norm problem but is not sparse:

x@arielo ~/thesis/notes/admm $ julia example.jl 
iter :         r     eps_pri           s    eps_dual
   1 :  1.07e+01    1.16e-01    5.33e+00    8.94e-03
  10 :  7.32e+00    1.31e-01    8.57e-01    7.99e-01
  20 :  3.71e+00    1.71e-01    6.30e-01    1.18e+00
  30 :  2.22e+00    1.73e-01    3.54e-01    1.31e+00
  40 :  1.48e+00    1.73e-01    2.24e-01    1.37e+00
  50 :  1.10e+00    1.71e-01    1.55e-01    1.41e+00
  60 :  8.57e-01    1.70e-01    1.16e-01    1.44e+00
  70 :  7.04e-01    1.69e-01    8.01e-02    1.45e+00
  80 :  5.90e-01    1.68e-01    6.79e-02    1.46e+00
  90 :  5.06e-01    1.68e-01    5.12e-02    1.47e+00
 100 :  4.40e-01    1.67e-01    4.43e-02    1.48e+00
 110 :  3.85e-01    1.67e-01    3.78e-02    1.49e+00
 120 :  3.39e-01    1.66e-01    3.20e-02    1.49e+00
 130 :  2.99e-01    1.66e-01    2.87e-02    1.50e+00
 140 :  2.69e-01    1.66e-01    2.47e-02    1.50e+00
 150 :  2.46e-01    1.66e-01    2.02e-02    1.51e+00
 160 :  2.29e-01    1.65e-01    1.73e-02    1.51e+00
 170 :  2.16e-01    1.65e-01    1.49e-02    1.51e+00
 180 :  2.03e-01    1.65e-01    1.50e-02    1.51e+00
 190 :  1.90e-01    1.65e-01    1.41e-02    1.52e+00
 200 :  1.80e-01    1.65e-01    1.15e-02    1.52e+00
 210 :  1.71e-01    1.65e-01    1.09e-02    1.52e+00
total iterations: 217
elapsed time: 42.691435953 seconds
B_cvx[1, 1:5]
12698429068239566e-32   .013939616755278827 -7353072051576453e-20   .06373070099702301  -1157651480454043e-30

B_julia[1, 1:5]
.0003370479204108641    .012555947526811096 -.019597468707497928    .022700579905230008 -.0013147338574289631

ERROR: assertion failed
 in include_from_node1 at loading.jl:120
while loading /home/x/thesis/notes/admm/example.jl, in expression starting on line 41

The code is here:

using MAT
using ParallelSparseRegression


function matrixlasso(dictionary, data, lambda=1)
    dimensions, components = size(dictionary)
    dimensions, samples = size(data)

    # [b1 b2]
    b = sparse(reshape(data, dimensions * samples))

    # [D 0]
    # [0 D]
    AA = kron(speye(samples), sparse(dictionary))

    # [c1 c2]
    x = lasso(AA, b, lambda)

    # [C]
    code = reshape(x, components, samples)
    return code
end


file = matopen("admm.mat")
A = read(file,"A")
B_cvx = read(file, "B")
X = read(file, "X")
cost = read(file, "cost")
close(file)

# ||AB - X||_2^2 + ||B||_1
B_julia = matrixlasso(A, X)

print("B_cvx[1, 1:5]","\n")
print(B_cvx[1, 1:5], "\n")
print("B_julia[1, 1:5]","\n")
print(B_julia[1,1:5], "\n")

assert(size(B_julia) == size(B_cvx))
assert(B_julia == B_cvx)

Data can be downloaded from:
http://ingenieroariel.com/static/admm.mat

[ingenieroariel]

Also did the following from Julia:

B_scipy = decomposition.sparse_encode(X', A', alpha=1)'
 @pyimport sklearn.decomposition as decomposition

And got the same result as matlab with CVX (best viewed directly in github):

julia> B_scipy = decomposition.sparse_encode(X', A', alpha=1)'
100x80 Array{Float64,2}:
  0.0          0.0139396   -7.35307e-5   0.0637307   0.0          0.197123      0.105221   …  -0.0853114   0.125546    -0.215341     0.00451901   0.16689      0.0         -0.162344 
  0.00671331  -0.271385     0.0         -0.0865927  -0.239533     0.0           0.130555      -0.247439   -0.192975     0.108291     0.129311    -0.00745488   0.151057     0.134188 
  0.092018    -0.0466571    0.0          0.0        -0.173526     0.0583904     0.0           -0.239976    0.0222884    0.0547041   -0.00838742  -0.0430508    0.0          0.0108664
  0.0          0.0406571    0.0710886   -0.0841806   0.149463    -0.213898     -0.155298       0.238002    0.0         -0.00231269   0.0654717   -0.30452      0.0429364   -0.181418 
 -0.222417     0.0         -0.0565775    0.0         0.0         -0.035914      0.0            0.302307    0.0205838    0.0845521   -0.21952     -0.383232     0.0         -0.107403 
 -0.143232     0.0736967   -0.0184351   -0.064996    0.0364668    0.0          -0.319453   …  -0.191406    0.0959937   -0.03323     -0.0321607    0.0          0.433653     0.138869 
 -0.0417249   -0.0537671   -0.129682     0.0         0.0         -0.0148316    -0.099432      -0.0530144  -0.202929    -0.131838     0.0424001   -0.152744     0.0525337    0.0      
  0.127261     0.11239      0.0          0.111226    0.0          0.026315      0.139523       0.101185   -0.229387     0.107155     0.0162594    0.189774    -0.205809     0.0790809
 -0.0288046    0.0          0.130798     0.0        -0.29608     -0.226412      0.0790443     -0.0115126   0.0         -0.368609     0.0         -0.0175567    0.104483    -0.0215514
  0.159154     0.187072     0.0         -0.0313609   0.0          0.0          -0.121287       0.0         0.0         -0.0390657    0.0          0.024583     0.228062     0.0      
  0.112463    -0.0586232   -0.0702961   -0.131088    0.11252      0.0          -0.125692   …  -0.0243822  -0.0109358    0.10058     -0.19238     -0.0525926    0.186771     0.0      
  0.0544893    0.0534809   -0.225345     0.151859   -0.00926664  -0.0113341     0.0           -0.0436159  -0.0338527    0.0229145    0.0108512    0.0          0.00151912   0.253073 
 -0.410077     0.0          0.0          0.0         0.0674603    0.0830968     0.14277        0.24455    -0.0582725   -0.421776    -0.0662308    0.0290147   -0.0463231   -0.169209 
  0.103649     0.0          0.0679544   -0.0282118   0.0834912    0.000702372   0.178511       0.254805    0.00543017   0.334662    -0.125654    -0.00758221   0.0915403    0.117483 
  0.17975     -0.147339     0.0          0.04693     0.184891    -0.0240954     0.184799       0.0397556   0.117944     0.172326     0.0          0.0121911    0.173312    -0.168928 
 -0.089977    -0.0446637    0.0715952    0.232478   -0.139833     0.0          -0.173907   …  -0.0382983  -0.0593752    0.125548    -0.00998831   0.0354989    0.0148703    0.0062226
  0.140243    -0.250244     0.288005    -0.168106   -0.31282     -0.179343      0.0855086      0.0911132   0.0975621    0.0         -0.111709     0.164241     0.111485     0.0      
  0.0849159    0.0         -0.0963193   -0.0744175   0.174541     0.0           0.136773       0.0        -0.0189869    0.0142278    0.0279454    0.0          0.0          0.0      
  0.0552466   -0.0134877   -0.14529      0.0521893   0.31103     -0.027733      0.0            0.0         0.301843     0.0609932    0.114953    -0.255255     0.306983     0.0      
 -0.0482881    0.00641886  -0.0462691    0.0337346   0.0211458    0.0736829     0.155617       0.088703    0.336031     0.28972     -0.015167     0.0308042    0.0412673   -0.152172 
  0.0          0.0225403    0.0          0.136902    0.10653     -0.202327      0.130218   …   0.0120922   0.0273708   -0.111827    -0.125273     0.118236     0.126329    -0.0746866
  0.0403421    0.0258216   -0.0884775    0.20906    -0.065059     0.022254      0.0870001      0.226462    0.23719      0.127652     0.0         -0.0496721    0.0          0.0      
  0.0          0.0          0.0850217   -0.171317    0.183628     0.101963      0.231549       0.0        -0.301888     0.115841    -0.167362     0.03563      0.124953     0.188337 
  0.0         -0.0394901    0.27371      0.0173336  -0.257816     0.161648      0.0            0.244729   -0.0502436    0.00320964  -0.0719819    0.0          0.0          0.0      
  0.0162171    0.251252     0.0729622    0.228595   -0.309659    -0.256088     -0.154805       0.256831    0.342689    -0.0034437    0.0732845   -0.04361      0.0         -0.0423222
  ⋮                                                               ⋮                        ⋱                            ⋮                                                            
  0.0          0.0         -0.135709     0.0         0.0155266    0.0          -0.119532      -0.016565   -0.211007     0.247943     0.0          0.0446025    0.0          0.0      
 -0.0542337    0.265554     0.0921252   -0.0476038   0.27764     -0.0682732    -0.0272853      0.234726   -0.0178333    0.0         -0.0248664    0.0         -0.325215    -0.086565 
  0.124329    -0.0025759    0.221306     0.0799693  -0.151647     0.0          -0.392836      -0.111459   -0.0358537    0.0780863    0.0          0.0352136    0.105638    -0.0119163
  0.100427     0.179975     0.0         -0.0844956   0.0142206    0.0           0.0628009      0.0         0.0453503   -0.0187895    0.0272987    0.0415041    0.0477133    0.0      
  0.101916     0.0         -0.0697573    0.0         0.107866     0.0303107     0.119486   …  -0.068872   -0.0398203    0.0952906    0.0          0.0          0.0         -0.265723 
  0.0163071    0.0         -0.133172     0.208391    0.154222     0.0           0.290136      -0.0216161  -0.0531057    0.0         -0.00708974  -0.0156749    0.0655515    0.0      
  0.0          0.0         -0.212975     0.0         0.0102621    0.0533782     0.0            0.0958267   0.0          0.0          0.0         -0.175486     0.0          0.0148493
 -0.291424     0.0837155    0.0          0.0857853   0.0          0.0307569     0.122385       0.0623261  -0.0420877    0.0          0.0153309    0.240045     0.127116     0.0      
  0.040017    -0.127659     0.0854545    0.0        -0.202186    -0.0432097     0.0327421      0.0        -0.0442597    0.0          0.0         -0.19467     -0.236476     0.0636748
  0.0286221   -0.0366417   -0.0990827   -0.0741034  -0.196447    -0.00559864   -0.0157355  …  -0.0164241   0.100856    -0.0513738   -0.197535     0.0689478    0.0          0.0      
 -0.0457259    0.303425     0.0          0.0        -0.0292353    0.0           0.229368       0.199846   -0.0529167    0.00215494  -0.290643     0.0         -0.03808     -0.0372205
 -0.206679    -0.15615     -0.0797169    0.120087   -0.215291     0.305089     -0.0241901     -0.233644   -0.0871398   -0.222093     0.20695      0.0559714   -0.0646555    0.19889  
 -0.148835     0.111808    -0.124896    -0.0817988   0.19326     -0.0489159    -0.0984961     -0.189682    0.0783978   -0.0290623   -0.00779019   0.217742    -0.118851     0.0729628
 -0.053811     0.0         -0.0211201    0.0         0.137885     0.0433281     0.114878       0.17886     0.0272867   -0.081256     0.0374359    0.113293     0.151196     0.0      
  0.0          0.231325    -0.0731003   -0.209764   -0.048146     0.0818697     0.0        …  -0.0569606   0.148913    -0.0860469    0.0         -0.0172812   -0.133877    -0.22088  
  0.15216      0.0          0.0489168    0.053746    0.0         -0.0235847     0.138471       0.0788577  -0.108394    -0.224967     0.072282     0.287115     0.0          0.0      
 -0.0901374   -0.185604     0.246231    -0.0941761  -0.269384    -0.316862     -0.341035      -0.537795   -0.129642    -0.0189861    0.0          0.0          0.0         -0.0427636
 -0.148731     0.0          0.00612961   0.0         0.0777175    0.0197545     0.0638228      0.132194   -0.01997     -0.0771284    0.125271    -0.147775     0.0919206    0.0929849
  0.176009    -0.027776     0.0442896    0.0         0.086653     0.0752426     0.220084       0.0380418  -0.0935554    0.0          0.0219196   -0.161433     0.106393     0.0      
  0.0          0.0          0.0476136   -0.003474   -0.161337     0.256054      0.0472292  …  -0.126704    0.10149      0.05294      0.159351     0.0         -0.316103     0.0      
 -0.0228349    0.0342999   -0.140112     0.0        -0.0733659    0.0           0.0802205      0.0732138   0.100556    -0.0304118    0.0         -0.090497     0.232528    -0.136202 
  0.0         -0.146676     0.0          0.0326132   0.173679    -0.0668674    -0.158378       0.0756563  -0.319589     0.0          0.135652     0.00611796  -0.00236722  -0.0476401
  0.00616579  -0.0550297    0.0145877   -0.11284     0.0         -0.0595453    -0.161158      -0.0929086  -0.470292     0.0146403    0.0         -0.194445     0.154948    -0.0403172
 -0.139795    -0.116678    -0.0975358    0.250031    0.0          0.0          -0.290872       0.256622    0.0         -0.39564     -0.0348308   -0.0226514    0.0          0.0  

julia> B_cvx
100x80 Array{Float64,2}:
  1.26984e-16   0.0139396    -7.35307e-5    0.0637307    -1.15765e-15   0.197123      0.105221     …   0.125546     -0.215341      0.00451901    0.16689      -4.34295e-14  -0.162344   
  0.00671331   -0.271385      1.60135e-12  -0.0865927    -0.239533      9.94123e-14   0.130555        -0.192975      0.108291      0.129311     -0.00745488    0.151057      0.134188   
  0.092018     -0.0466571    -5.25166e-13  -4.64765e-11  -0.173526      0.0583904    -8.4439e-17       0.0222884     0.0547041    -0.00838742   -0.0430508     5.31588e-14   0.0108664  
  1.07977e-14   0.0406571     0.0710886    -0.0841806     0.149463     -0.213898     -0.155298         2.46853e-16  -0.00231269    0.0654717    -0.30452       0.0429364    -0.181418   
 -0.222417      3.42479e-12  -0.0565775     5.22998e-12   1.86998e-16  -0.035914     -1.37868e-16      0.0205838     0.0845521    -0.21952      -0.383232     -1.62462e-14  -0.107403   
 -0.143232      0.0736967    -0.0184351    -0.064996      0.0364668    -7.70142e-14  -0.319453     …   0.0959937    -0.03323      -0.0321607    -2.64668e-12   0.433653      0.138869   
 -0.0417249    -0.0537671    -0.129682      4.31252e-12  -4.689e-16    -0.0148316    -0.099432        -0.202929     -0.131838      0.0424001    -0.152744      0.0525337    -3.80182e-12
  0.127261      0.11239      -1.06557e-12   0.111226      1.61246e-15   0.026315      0.139523        -0.229387      0.107155      0.0162594     0.189774     -0.205809      0.0790809  
 -0.0288046    -2.44397e-11   0.130798      3.83147e-11  -0.29608      -0.226412      0.0790443       -1.22876e-16  -0.368609     -1.45545e-13  -0.0175567     0.104483     -0.0215514  
  0.159154      0.187072     -1.51861e-12  -0.0313609     1.50293e-16   5.54167e-14  -0.121287         4.01235e-17  -0.0390657     4.3033e-13    0.024583      0.228062      7.16578e-12
  0.112463     -0.0586232    -0.0702961    -0.131088      0.11252      -2.56007e-13  -0.125692     …  -0.0109358     0.10058      -0.19238      -0.0525926     0.186771      2.6878e-11 
  0.0544893     0.0534809    -0.225345      0.151859     -0.00926664   -0.0113341    -5.02133e-17     -0.0338527     0.0229145     0.0108512    -1.06462e-13   0.00151912    0.253073   
 -0.410077      1.11298e-11  -4.43362e-12  -6.75429e-11   0.0674603     0.0830968     0.14277         -0.0582725    -0.421776     -0.0662308     0.0290147    -0.0463231    -0.169209   
  0.103649      1.08989e-11   0.0679544    -0.0282118     0.0834912     0.000702372   0.178511         0.00543017    0.334662     -0.125654     -0.00758221    0.0915403     0.117483   
  0.17975      -0.147339      1.75041e-12   0.04693       0.184891     -0.0240954     0.184799         0.117944      0.172326      6.13239e-13   0.0121911     0.173312     -0.168928   
 -0.089977     -0.0446637     0.0715952     0.232478     -0.139833      1.84145e-14  -0.173907     …  -0.0593752     0.125548     -0.00998831    0.0354989     0.0148703     0.0062226  
  0.140243     -0.250244      0.288005     -0.168106     -0.31282      -0.179343      0.0855086        0.0975621    -1.06792e-16  -0.111709      0.164241      0.111485     -2.81678e-11
  0.0849159     2.44908e-11  -0.0963193    -0.0744175     0.174541     -1.07807e-13   0.136773        -0.0189869     0.0142278     0.0279454    -2.51349e-12  -2.56534e-14   1.49469e-11
  0.0552466    -0.0134877    -0.14529       0.0521893     0.31103      -0.027733      1.36437e-16      0.301843      0.0609932     0.114953     -0.255255      0.306983     -2.8078e-11 
 -0.0482881     0.00641886   -0.0462691     0.0337346     0.0211458     0.0736829     0.155617         0.336031      0.28972      -0.015167      0.0308042     0.0412673    -0.152172   
 -1.97109e-15   0.0225403    -7.01816e-13   0.136902      0.10653      -0.202327      0.130218     …   0.0273708    -0.111827     -0.125273      0.118236      0.126329     -0.0746866  
  0.0403421     0.0258216    -0.0884775     0.20906      -0.065059      0.022254      0.0870001        0.23719       0.127652     -4.13464e-13  -0.0496721     6.55799e-14  -1.02061e-11
  2.15158e-14  -2.59867e-11   0.0850217    -0.171317      0.183628      0.101963      0.231549        -0.301888      0.115841     -0.167362      0.03563       0.124953      0.188337   
 -2.12543e-15  -0.0394901     0.27371       0.0173336    -0.257816      0.161648     -2.90586e-16     -0.0502436     0.00320964   -0.0719819    -3.28179e-13  -3.18788e-14  -5.27976e-13
  0.0162171     0.251252      0.0729622     0.228595     -0.309659     -0.256088     -0.154805         0.342689     -0.0034437     0.0732845    -0.04361      -4.16791e-14  -0.0423222  
  ⋮                                                                     ⋮                          ⋱                 ⋮                                                                  
  3.02952e-14   4.48974e-12  -0.135709     -4.65853e-12   0.0155266     1.4682e-13   -0.119532        -0.211007      0.247943      6.57317e-14   0.0446025     6.37175e-14  -2.72854e-11
 -0.0542337     0.265554      0.0921252    -0.0476038     0.27764      -0.0682732    -0.0272853       -0.0178333     1.13075e-17  -0.0248664    -1.9957e-12   -0.325215     -0.086565   
  0.124329     -0.0025759     0.221306      0.0799693    -0.151647      1.22538e-13  -0.392836        -0.0358537     0.0780863    -6.14396e-13   0.0352136     0.105638     -0.0119163  
  0.100427      0.179975     -2.39445e-12  -0.0844956     0.0142206    -2.65506e-13   0.0628009        0.0453503    -0.0187895     0.0272987     0.0415041     0.0477133     3.81477e-11
  0.101916     -4.92003e-12  -0.0697573     5.75295e-11   0.107866      0.0303107     0.119486     …  -0.0398203     0.0952906    -8.66129e-13   3.70512e-12  -4.65128e-14  -0.265723   
  0.0163071    -2.78817e-11  -0.133172      0.208391      0.154222      1.99108e-13   0.290136        -0.0531057    -1.50316e-16  -0.00708974   -0.0156749     0.0655515    -6.48394e-12
  8.16982e-15  -6.81406e-12  -0.212975      4.73686e-12   0.0102621     0.0533782    -1.17497e-16      2.2312e-16    1.78399e-16   2.89511e-13  -0.175486      1.92221e-14   0.0148493  
 -0.291424      0.0837155     2.77985e-12   0.0857853    -1.70895e-16   0.0307569     0.122385        -0.0420877     1.82848e-17   0.0153309     0.240045      0.127116      5.63598e-12
  0.040017     -0.127659      0.0854545     1.85898e-11  -0.202186     -0.0432097     0.0327421       -0.0442597     7.47656e-17   8.09253e-15  -0.19467      -0.236476      0.0636748  
  0.0286221    -0.0366417    -0.0990827    -0.0741034    -0.196447     -0.00559864   -0.0157355    …   0.100856     -0.0513738    -0.197535      0.0689478     2.54892e-14  -3.57924e-11
 -0.0457259     0.303425      3.06233e-12   2.76056e-11  -0.0292353     8.33311e-14   0.229368        -0.0529167     0.00215494   -0.290643     -5.10446e-14  -0.03808      -0.0372205  
 -0.206679     -0.15615      -0.0797169     0.120087     -0.215291      0.305089     -0.0241901       -0.0871398    -0.222093      0.20695       0.0559714    -0.0646555     0.19889    
 -0.148835      0.111808     -0.124896     -0.0817988     0.19326      -0.0489159    -0.0984961        0.0783978    -0.0290623    -0.00779019    0.217742     -0.118851      0.0729628  
 -0.053811      1.86443e-11  -0.0211201     2.89916e-11   0.137885      0.0433281     0.114878         0.0272867    -0.081256      0.0374359     0.113293      0.151196     -9.44555e-12
  1.45399e-14   0.231325     -0.0731003    -0.209764     -0.048146      0.0818697     3.04947e-16  …   0.148913     -0.0860469     4.31955e-13  -0.0172812    -0.133877     -0.22088    
  0.15216       3.70605e-12   0.0489168     0.053746      1.94918e-16  -0.0235847     0.138471        -0.108394     -0.224967      0.072282      0.287115      1.72256e-14   7.90182e-12
 -0.0901374    -0.185604      0.246231     -0.0941761    -0.269384     -0.316862     -0.341035        -0.129642     -0.0189861     9.34785e-14   2.55009e-12  -3.30011e-14  -0.0427636  
 -0.148731     -7.95238e-12   0.00612961    1.63064e-11   0.0777175     0.0197545     0.0638228       -0.01997      -0.0771284     0.125271     -0.147775      0.0919206     0.0929849  
  0.176009     -0.027776      0.0442896    -3.98776e-11   0.086653      0.0752426     0.220084        -0.0935554     7.99783e-17   0.0219196    -0.161433      0.106393     -1.16811e-11
  4.05117e-14  -9.22803e-12   0.0476136    -0.003474     -0.161337      0.256054      0.0472292    …   0.10149       0.05294       0.159351     -2.02335e-12  -0.316103     -1.74537e-11
 -0.0228349     0.0342999    -0.140112      2.21944e-11  -0.0733659    -1.06882e-13   0.0802205        0.100556     -0.0304118     4.69221e-13  -0.090497      0.232528     -0.136202   
 -4.10374e-15  -0.146676     -3.98025e-13   0.0326132     0.173679     -0.0668674    -0.158378        -0.319589     -9.85169e-17   0.135652      0.00611796   -0.00236722   -0.0476401  
  0.00616579   -0.0550297     0.0145877    -0.11284      -8.54843e-16  -0.0595453    -0.161158        -0.470292      0.0146403     2.48209e-13  -0.194445      0.154948     -0.0403172  
 -0.139795     -0.116678     -0.0975358     0.250031      6.22888e-16  -3.04969e-14  -0.290872        -1.08957e-16  -0.39564      -0.0348308    -0.0226514    -1.97114e-15  -3.90435e-12

julia> B_julia
100x80 Array{Float64,2}:
  0.000337048   0.0125559    -0.0195975     0.0227006    -0.00131473    0.158092      0.000663977  …   0.00132733   -0.0556719     0.0859936     0.0901456     0.0274917    -0.156614   
 -0.000999775  -0.137709     -0.00189461   -0.108788     -0.133136      0.10675       0.0133637       -0.286436     -0.0329385    -0.00322267    0.000318102  -0.135325      0.00885051 
  0.0737673    -0.00030499   -0.00249059    0.0261302     0.124452      0.0320024    -0.000321017      0.000889766  -0.000529372   0.119507     -0.00173593    0.000114643  -0.00306437 
 -0.079319      0.000324243  -0.00339591   -0.00036351    0.000527439   0.00130374   -0.0467694        0.000461938  -0.0609708     0.0690949    -0.0704039    -0.00167842   -0.00103464 
 -0.102341     -0.0749729     0.00371976    0.0016458     0.00128979   -0.000929159   0.000943008      0.00044275    0.0779099    -0.000608869  -0.000409666   0.0311462    -0.000302858
 -0.120621     -0.000311348  -0.0655874    -0.000266848   0.145086      0.0383281    -0.125069     …  -0.00120831   -0.108513      0.00230788    0.0651062     0.241779      0.075382   
  0.0758896    -0.00591877   -0.0606568    -0.000956742   0.0374417     0.0186736    -0.232604        -7.72066e-5    0.0481777     0.0366206    -0.139991     -0.000495568  -0.00219826 
  2.88976e-6    0.0758953    -0.003234      0.146643     -0.00052971    0.00120368   -0.00080923       0.0969681     0.159552      0.0429008     0.182256     -0.0667012    -3.46449e-5 
 -0.0019605    -0.000235125  -0.000844206   0.0010696    -8.86654e-5   -0.000233001  -0.092073        -0.000446137  -0.108565      0.00243809   -0.0174647     0.079925     -0.0865091  
  0.0781275     0.19871      -0.00191063   -0.000804741   0.199612     -0.104539     -0.0670915       -0.000663136  -0.000714714   0.0028652     0.0371377     0.0430611     0.000523397
  0.0158877    -0.000120918  -0.067413     -0.00103618    0.0422939     0.0367469    -0.184133     …   0.0103255    -0.00105483   -0.227575     -0.000630592   0.0127628     0.000649823
  0.0259366    -0.0523148    -0.213944      0.00967588   -0.000993193  -0.00093327    0.0600554       -0.000914976   0.0999781     0.010863      0.000491431   0.0744603     0.141033   
 -0.234498      4.58081e-5   -0.00417361   -0.00153287    0.0121918     0.0353958     0.0706351       -0.0849251    -0.337175     -0.056515      0.0388506     0.0619645    -0.14783    
 -0.000448261  -0.000955057   0.00174637   -0.113326      0.000128093  -0.00111729    5.39709e-5       0.000604747   8.90301e-5    0.00125488    0.000862802   0.000424115   0.00045246 
  0.000258625   0.00113976    0.00281135    0.000888562   0.000153247  -0.000210421   0.0162539        0.0725167    -0.000309987  -0.0835678    -0.000392986  -0.000816783  -0.0431199  
 -0.0234674     0.113753      0.0023156     0.0505136    -0.000580892   0.000614167   0.000960195  …  -0.0645927    -0.000302794   0.000845978  -0.00129536    3.43712e-5   -0.00010349 
  0.109872     -0.0743854     0.229784     -0.230191      0.0417266    -0.143704     -0.0520787       -0.00120834    0.130588     -0.16166      -0.000530155   0.13643      -0.0716151  
  0.0289604     0.0916581     0.00287153   -0.013683      0.000441428  -0.000266201  -0.000283894     -2.78584e-6    0.0423694     0.0895212     0.000583212   0.000901425  -0.0001322  
  0.0778609    -0.185006     -0.147858     -0.115069      0.191568      0.000501431  -0.191749         0.170681      0.0887981     0.108111     -0.102144      0.00102936    0.00109931 
 -0.0704627     0.0072464    -0.00162096   -0.000514679  -0.0010019     0.126287     -0.00097227       0.0899665     0.000374993   0.041103      0.0919705    -0.00415496   -0.0223385  
  0.000400198   0.000269394  -0.00628424    0.221756     -0.000830738  -0.119343      0.010114     …   0.000188966  -0.0699956    -0.0981138     0.0646589     0.00070182   -0.0386415  
 -8.73404e-5   -0.000224535   0.00271826    0.0796657    -0.169764      0.020407      0.0182277       -0.00018863    0.0903883    -0.134761     -0.116892      0.000953686  -5.07551e-5 
  0.0313879     0.00290555   -0.00222414    0.00126002    0.000427026   0.0152981     0.0370784       -0.139747      0.00259641    0.00189545    0.000599819   0.0248655     0.0324957  
 -0.013871     -0.000992704   0.0263489     0.000919314  -0.192936      0.0320112    -9.01476e-5       0.0799235     0.000481464  -0.00210076   -0.123522     -0.173831     -0.00067874 
  0.00112559    0.256739      0.00153983    0.0237268    -0.0200846    -0.212301     -0.000243966      0.0113312     0.00198908   -0.0483714     0.166568     -0.0800781     0.0669477  
  ⋮                                                                     ⋮                          ⋱                 ⋮                                                                  
  0.00092431    0.13294      -0.166694      0.0828401     0.012587      0.00164478    0.00079533      -0.205943      0.02711      -0.0647445    -0.00161003    0.086439      0.2168     
  0.000602251   0.000274472  -0.00135318    0.0529948     0.0425002     0.000466094   0.00120179       0.000596479   0.00298202   -0.00117354    0.00197412   -0.145644     -0.00112347 
  0.0954518    -0.0126554    -0.0263831     0.0956105    -0.126602      0.112251      0.000901736      0.00134231    0.029124     -0.000320644  -0.000342259  -0.00169632   -0.00392543 
 -0.00101901    0.191769      0.0345135    -5.62647e-5    0.000871625  -0.000171948   0.00013333      -0.079211     -0.0189447     0.00312025    0.00041664    0.224698      0.0637783  
  0.164225      0.0761533     0.00241984    0.0436349    -0.000244279  -0.0383924    -0.0492106    …  -0.0581738     0.0620536     0.0056159    -0.000320606   0.00363909   -0.256293   
  0.0145109     0.00125964   -0.0531116     0.16977      -0.00106843   -0.0530862     0.0701563       -0.0770485    -0.119355      0.00166972    0.0151789     0.0609907     0.0988225  
 -0.000429347  -0.000298952  -0.0899243    -0.0719078     0.151216     -0.000614316   0.00164534       0.000967415  -0.0591879    -0.125367     -0.100942     -0.00146107    0.000989065
 -0.250254      0.0928965     0.00332207    0.00121716    0.0526083     0.0986189     0.198085        -0.0568833    -0.00185639    0.0932408     0.135025      0.017814      0.00343379 
  0.00740545   -0.128521      0.0737627    -0.00571919   -0.0942927     0.00261609    0.161746        -0.0116408    -0.00133246    0.0017285    -0.111471     -0.154944     -0.0217122  
 -0.000599954  -0.0902916    -0.0761979    -0.081075     -0.126232     -0.000427275   0.101758     …   0.0729002     0.00337086   -0.112974      0.0821665    -0.00426502   -0.0340602  
  0.0767949     0.26242      -0.0138063    -0.000765209  -0.19055      -0.0332927    -0.000556921     -0.00172655   -0.0177578    -0.00599413    0.0422183     0.143861      4.28319e-5 
 -0.278005      0.00037781   -0.0749007     0.080908      0.00112031    0.279345      0.137768         0.0304783    -0.101874      0.106087      0.000860824  -0.00314154   -0.0811042  
 -0.000280719  -0.167395     -0.162634     -0.0898738     0.0242555    -0.000327285   0.000237801     -0.0586578     0.0170453    -0.00167834   -3.20847e-5   -0.0179528     0.000424374
  0.149003      0.000501913  -0.00325152   -0.000169014  -0.000717131  -0.125752      0.000504163      0.054423      0.000501139   0.00300892    0.080138      0.135118     -0.0759846  
 -0.000314447   0.000533176  -0.0012193    -0.0250234     0.00113922    0.0621237     0.0959975    …  -0.000303605  -0.00126875    0.000462062  -0.00112783    0.000763351  -0.000826756
 -0.000824322  -0.000311654   0.00119355    0.0229343    -0.0382838    -0.13775       0.111224        -7.36839e-5   -0.0811097    -0.00190996    0.0800673    -0.103401     -0.000566397
 -0.00101375   -0.13783      -0.00190418    0.000867497  -0.0374163    -0.00158244   -0.0946434        0.0556102    -0.00230705    0.149802     -0.000523509   0.000788743  -0.117547   
 -0.0857366    -0.146055      0.00321863    0.024037     -0.0126447     0.127917      0.207326         0.00102758   -0.0504689     0.194544     -0.0378578    -0.00212678    0.147205   
  0.13645      -0.0844313     0.198377     -0.0512691     0.00132027    0.141729      0.000561085     -0.000128916   0.000250492   0.0496934    -0.218118      0.00396689   -0.00120344 
 -0.00117122   -0.0744178     0.0690885     0.000536034  -6.27491e-5    0.000557838  -0.01064      …  -0.000698088  -0.00197782    0.00121599   -0.16723      -0.304525      0.00524283 
  0.0198386    -0.000519034  -0.213309     -0.0247626     0.00127581   -4.09628e-5    0.063534        -0.00132398    0.0320213    -0.107832     -0.0414453     0.0748763    -0.179304   
 -0.00268172   -0.00194285    0.203731     -0.000252043  -0.00012673   -0.198046     -9.41217e-5      -0.0808084     0.000409373   0.00194694    0.000247766  -0.0010174    -0.000681956
 -0.0477631     0.107093      0.000765209  -0.000792182   0.0941726     0.0180149     0.000616269     -0.312714      0.000618501   0.0004254    -0.000130796   0.228201      0.00308488 
  0.0589122     0.00122412   -0.21586       0.000635813   0.000404807   0.000516293  -0.0501303        0.000608686  -0.103942      0.00234117    0.000168612   0.108051      0.000175169