CK repository to reproduce our recent papers

Shared artifacts in the Collective Knowledge Format with JSON API and JSON meta information to reproduce and extend techniques from the following papers on "Collective Mind" and "Collective Knowledge":

• http://arxiv.org/abs/1506.06256
• http://hal.inria.fr/hal-01054763
• https://hal.inria.fr/inria-00436029
• http://arxiv.org/abs/1407.4075

Status

This is more a proof-of-concept repository. You can find reproducible workflows and articles in the CK format from the latest computer systems' conferences here.

For example, check out this reusable and customizable artifact from CGO'17 with automatic cross-platform software installation and web-based experimental dashboard powered by the CK framework:

• GitHub CK repo
• Paper with artifact appendix
• PDF snapshot of the interactive CK dashboard
• CK concepts

Authors

• Grigori Fursin, dividiti/cTuning foundation
• Anton Lokhmotov, dividiti

Prerequisites

• Collective Knowledge Framework

Installation

Compatibility

• Linux, Windows, Android

Dependencies:

• CK: http://github.com/ctuning/ck
• Various python packages on Linux: sudo apt-get install python-numpy python-scipy python-matplotlib

ck pull repo:reproduce-ck-paper

Usage

##########################################################

Reproducible experiments (collaboratively validate assumption and report unexpected behavior for further analysis):

1. Validating that threshold filter needs run-time adaptation to achieve best performance and energy usage depending on the type of image:

To reproduce/validate expectation:

cd script/reproduce-filter-speedup

python reproduce.py

From ipython notebook (check that CK is installed as python module):

If (un)expected behavior is reported (considerable slow downs or speedups for the first 2 optimizations), you will be asked to share results in a public cknowledge.org/repo to demonstrate crowdsourcing of experimentation and validation of results.

You can later view shared results at http://cknowledge.org/repo/web.php?wcid=bc0409fb61f0aa82:8404df882462f978&subview=reproduce-ck-paper-filter

cp reproduce.ipynb.remove_this_extension reproduce.ipynb ipython notebook reproduce.ipynb

########################################################## 2) Analysis of variation of experimental results (for the same experiment) - density analysis and peak detection can provide expected values and missing features.

To reproduce/validate expecation (ck should be installed as Python package using "ck setup kernel --install):

cd script/reproduce-filter-variation

python reproduce.py

Feedback

• https://groups.google.com/forum/#!forum/collective-knowledge