/GSDEstimator

Estimator of Generalized Score Distribution's parameters using the OpenCL library.

Primary LanguageC++

GSD Estimator

Estimator for Generalized Score Distibution's parameters using OpenCL for accelerating calculation on GPI.

Build process

In case of doubts refere to wiki page.

Requirements

For windows you'll need Visual Studio Build Tools. For linux and Mac OS X just these g++ make cmake should be enough.

Note: When cloning this repo, please make sure to add the --recursive flag to your Git call.

Note for Mac OS X users: You can install gcc through Homebrew.

Note for M1 CPU users: The code will not work on your machine. 😅

Command sequence

mkdir build
cmake -B build
cmake --build build --config Release

Note for Mac OS X users: cmake will, by default, use the AppleClang compiler. You need to manually instruct it to use the GCC compiler instead. To do so, prepend all calls to cmake with the following:

CC=/opt/homebrew/Cellar/gcc/11.1.0_1/bin/gcc-11 CXX=/opt/homebrew/Cellar/gcc/11.1.0_1/bin/g++-11

(Make sure to adapt the line above according to your GCC installation.)

For linux the final executable named GSDEsimator will end up in build directory. For windows you will find it in build/Release directory.

Usage

Usage: ./GSDEstimator <samples_file> [output_file]
	samples_file: name of input file containg samples for estimation
	output_file: name of csv file containing results
  • samples file is a text file with each sample in new line. Every sample consits of 5 values from subjective experiments utilizing the Absolute Category Rating (ACR) technique, with possible answers {1,2,⋯,5}
  • output file will be a csv file with results decribed by values
    • idx - index of sample
    • psi - estimated mean (rating)
    • rho - estimated variance
    • log_likelihood - logarithm of likelihood value