Release 0.20

Installing libskylark

There are two options for quickly installing libskylark in Linux-x86_64 systems (i.e. Linux distribution independent):

• From conda channels (full installation, including FFTW-based sketching)
• From a single-file installer (not including FFTW-based sketching)

Installing libskylark from conda channels

Open a terminal and type (copy-paste):

# Get miniconda for python 2.7
curl -L -O https://repo.continuum.io/miniconda/Miniconda2-latest-Linux-x86_64.sh
bash Miniconda2-latest-Linux-x86_64.sh -b -p ${HOME}/miniconda2

# Get the libskylark channel
curl -L -O https://github.com/xdata-skylark/libskylark/releases/download/v0.20/channel.tar.gz
cd ${HOME}; tar -xvzf channel.tar.gz

# Install fftw and libskylark conda packages
export PATH=${HOME}/miniconda2/bin:${PATH}
conda install fftw --yes --channel conda-forge
conda install libskylark --yes --channel file://${HOME}/channel

Trying the installation

Start an ipython shell:

ipython

and then input the following python snippet:

import El
from skylark import sketch
A = El.DistMatrix()
El.Uniform(A, 10, 10)
S = sketch.FJLT(10, 4) 
SA = S * A
print [[SA.Get(i, j) for j in range(10)] for i in range(4)]

Congratulations! You have just sketched a 10x10 distributed matrix of uniform distribution random entries into a 4x10 sketched matrix by making use of the Fast Johnson-Lindenstrauss Transform (FJLT).

Now consider hard-wiring the installation path by appending this line

export PATH=${HOME}/miniconda2/bin:${PATH}

to your ~/.bashrc.

Installing libskylark from a single-file installer

Open a terminal and type (copy-paste):

# Download the installer and install libskylark
curl -L -O https://github.com/xdata-skylark/libskylark/releases/download/v0.20/install.sh
bash install.sh -b -p ${HOME}/libskylark
export PATH=${HOME}/libskylark/bin:${PATH}

Trying the installation

Start an ipython shell:

ipython

and then input the following python snippet:

import El
from skylark import sketch
A = El.Matrix()
El.Uniform(A, 10, 10)
S = sketch.JLT(10, 4) 
SA = S * A
print SA.ToNumPy()

Congratulations! You have just sketched a 10x10 matrix of uniform distribution random entries into a 4x10 sketched matrix by making use of the Johnson-Lindenstrauss Transform (JLT).

Now consider hard-wiring the installation path by appending this line

export PATH=${HOME}/libskylark/bin:${PATH}

to your ~/.bashrc.

Happy sketching!