ia080-classifiers-comparison's Introduction

IA080-classifiers-comparison

IA080 Knowledge discovery seminar

Quick guide

$ cd data/extra-data/ && unzip *.zip && cd ../..
$ cd data/metal-data/ && unzip *.zip && cd ../..
$ cd data/weapon-data/ && bash download_data.sh && unzip *.zip && cd ../..
$
$ python3 -m venv venv
$ . venv/bin/activate
$ pip install -r requirements.txt
$ cd src
$ python runner.py 2>/dev/null > output/my_output.csv

$ nice -n 15 taskset -c 5-6 \time -v -o time_stats.txt python runnner.py > output/aura_output.csv

XCDG data generator

In the /generator folder you can find the XCDG.jar along with example input files used for the creation of data in /extra-data. The generator can be called from the command line with

java -jar XCDG.jar input.xml output.data

Or you can output the data in the .csv or .arff format. For further information on XCDG, please refer to the corresponding bachelor thesis, namely chapter 5.

ia080-classifiers-comparison's People

Contributors

Watchers

ia080-classifiers-comparison's Issues

Timeout for fitting process

Add support for a timeout to classifiers fitting process.

Extreme Learning Machine

Extreme Learning Machine
(Wiki, article)

While this architecture may not outperform other NN types in terms of classificaton accuracy, the learning speed to accuracy ratio should make ELM competitive both with NNs and classical ML methods.

Before creating the datasets for specific situations, I wanted to discuss the size of the data, i.e. rough number of examples and attributes so that the learning won't take ages for each of them while being able to train the networks properly.

Also, if you have an idea for a particular dataset, please let me know.