• Brief Description
• Compilation and Installation
• UCI Examples

This is the MCP system designed to transform Big Data into propositional formulas. It consists of several modules. The main part is the MCP core. In addition to the core there are prequel and sequel supporting modules.

MCP core has four variants:

mcp-seq     (sequential),
mcp-mpi     (parallel using the Message Passing Interface),
mcp-pthread (parallel using POSIX threads), and
mcp-hybrid  (parallel using a combination of MPI and POSIX threads).

The prequel modules are

mcp-guess (for guessing the structure of the input data),
mcp-trans (for binarization of input data), and
mcp-split (for splitting binarized data into a learning and checking part)

The sequel modules are

mcp-check (for checking the accuracy of the produced formula)

Additionally, the distribution contains the following files:

README.md (this document),
LICENCE   (GNU General Public Licence v.3, under which this software is distributed),
Makefile  (for the compilation and installation of the MCP system)

• Outline of the distribution
• Compilation
• Invocation

The subdirectories of this distribution are:

• src containing the sources of the MCP system,
• man containing the manual pages,
• paper containing the PDF document mcp-sat.pdf with a detailed description of the MCP system
• bin is necessary for compilation, and
• uci containing several examples from the UCI Machine Learning Repository, treated by the MCP system.

A C++ compiler satisfying at least the C++17 revision is necessary to successfully compile the MCP system, mainly for the use of auto types and POSIX threads. Only the standard library is used, therefore there is no need to install any additional C++ libraries. The g++ GNU Project compiler is used in the Makefile. If you have a different compiler, please modify the Makefile according to your installation.

You need to have installed the Message Passing Interface (MPI) to be able to compile the modules mcp-mpi and mcp-hybrid. However, this is not an absolute necessity. The MCP system functions even without this two variants of the MCP core.

To compile and install the MCP system, write the command

   make

in the root directory of this unpacked tarball. The system then

• compiles the sources,
• places the binaries in the root directory of the installation,
• installs the manual pages in the directories /usr/local/share/man/man1 and /usr/local/share/man/man5,
• installs the binaries in the directory /usr/local/bin.

You need to have superuser privileges to execute the last two parts.

If you do not have MPI installed, compile and install the MCP system (without the modules mcp-mpi and mcp-hybrid) by the command

    make no-mpi

The uci subdirectory contains the following examples from the UCI Machine Learning Repository, prepared to be treated by the MCP system. These examples are

• abalone (identifying abalone with 27 rings),
• balance-scale (identifying psychological experiments balancing a scale),
• balloons (a toy example, where specific formulas are required to be produced),
• breast-cancer-wisconsin (identifying benign and malignant breast cancer cases in Wisconsin),
• car (identifying very good cars),
• forest-fire (predicting forest fires in July, August, and September),
• iris (identifying three types of iris flowers),
• mushroom (identifying edible and poisonous mushrooms), and
• vote (identifying democrats and republicans in the House of Representatives according to the 1984 US Congressional Voting Records).

All examples are equipped with a Makefile, facilitating the execution of the MCP system on them. Some of these examples use the parallel modules mcp-mpi or mcp-hybrid. If you did not install them, please replace the commands mpirun mcp-mpi and mpirun mcp-hybrid by mcp-pthread.

EOF