This repository contains all the code implementing the Framework SapoSVM as described in the manuscript

• A. Alves, J.O.L. Silva, L.C. Matioli, P.S.M. Santos. "A new framework for training Support Vector Machines".

If you run into problems please do not hesitate to contact me:

• Jonatas O. L. Silva [email protected]

Directory organization

• code/ : Directory containing the methods implementation necessary to run all experiments

• data/ : Directory containing the datasets used in the experiments

• results/ : Directory which will stores the results after run the file run_results.py

• main_experiment_performance.m : Run this to generate inside of the folder results/ the txt files necessary to plot the performance profile according to the Section 4.2 of the paper

• main_experiment_qc.m : Run this to generate inside of the folder results/ the txt file responsible to generate the experiments using the quasi-Couchy approximation in step 1 of the framework

• main_experiment_sp.m : Run this to generate inside of the folder results/ the txt file responsible to generate the experiments using the spherical approximation in step 1 of the framework

• plot_main_performance.m.m : Run this to plot the profile performance according to the txt file in results/ folder

• run_results.py : Run this to generate the .tex file inside of results/ folder

Obs: run the file run_results.py only after generate the .txt files using main_experiment_qc.m, main_experiment_qn.m and main_experiment_sp.m

• summarizy.py : Code in Python with scripts used to generate the .tex file

Generating table results

To generate Tables 5, 6 and 7 presented in the paper please follow the steps below:

1 - Open the Matlab and run the files: main_experiment_qn.m, main_experiment_qc.m and main_experiment_sp.m. After it three .txt files will be generated. The process can during some minutes.

2 - Run the command below:

  python run_results.py

Then, inside of the folder results/ a file .tex will be created containing the three tables presented in the paper

Generating performance profile

1 - Open the Matlab and run the file main_experiment_performance.m. After it a .txt file will be generated.

2 - Open the Matlab and run the file plot_main_performance.m. After it three figures will be generated according to the paper.

Dependencies

Instal Libsvm package according to the official documentation:

https://www.csie.ntu.edu.tw/~cjlin/libsvm/