Manpages

This page contains all the manual pages of the AI model selection toolset. Note that the following manpages can also be viewed from a terminal, either using a precompiled version or by compiling them manually using rst2man MANPAGE and then using its output as the input of the man command.

ai-models

train and test ai models for UX-SAD

Manual section

1

Manual group

UX-SAD

Author

Andrea Esposito <a.esposito39@studenti.uniba.it>

Version

1.0.0

Synopsis

ai-models [-r RANDOM] [-o OUT] [-V] [-P] [-j JOBS] [-m MODEL] [-s STRATEGY]… DATASET EMOTION WIDTH HALF

ai-models [-h|-v]

Description

Train and test various AI models using multiple algorithms. This can be used either to train a specific model using a specific algorithm on a specific dataset or to compare the performances of multiple models.

The training uses the dataset provided in the folder DATASET (see the section The Dataset for more information on the dataset format) and will train a model to predict the emotion EMOTION using the half HALF of windows of WIDTH milliseconds.

The accepted value for the positional arguments are the following:

DATASET

Any path to a folder. The content of the folder must follow the specification provided in the section The Dataset.

EMOTION

The available emotions depend on the provided dataset. The standard ones are the “universal” emotions identified by P. Ekman plus two additional values, ie. EMOTION must be one in {anger|contempt|disgust|engagement|fear|joy|sadness|surprise|users|valence|websites}

WIDTH

The width of the window frames to be considered. The available values depend on the actual dataset.

HALF

The window half to be considered. Must be one of {before|after|full}

The available options are the following. Mandatory arguments to long options are mandatory for short options too.

-h, –help

Show this message and exit.

-v, –version

Show the version and exit.

-r, –random RANDOM

Set the random seed to RANDOM

-o, –out OUT

Set the path where the model will be saved to OUT. A new file with the same name as the model’s emotion will be created inside this directory.

-V, –verbose

If specified, increase the tool verbosity (messages will be printed to stderr).

-P, –progress

If specified, show the progress of the tool using a loading bar.

-j, –jobs JOBS

Set the number of parallel jobs.

-a, –algorithm {forest|svm|adaboost|tree|perceptron}

Set the algorithm on which the model will be based.

-s, –strategy [sfs|pca|efs]

The strategy to apply in order to reduce the number of final features.

The Dataset

The dataset specified by the DATASET argument must follow a specific structure. This is automatically done by the provided tools (see dataset.py), but if manual processing is done be sure to adhere the this assumptions before running the tool (otherwise, the outcome is unpredicted).

The folder DATASET must contain multiple CSV files: one for each emotion. The names of this files must be exactly “emotion.csv”, where emotion is the name of the emotion in lower case. Each CSV file must contain an header row. Each CSV must contain the keys middle.emotions.emotion, as well as each of the other emotions (that will be discarded by the tool once the dataset has been loaded).

classify

convert continous emotion values to discrete classes

Manual section

1

Manual group

UX-SAD

Author

Andrea Esposito <a.esposito39@studenti.uniba.it>

Version

1.0.0

Synopsis

classify <INPUT_FILE >OUTPUT_FILE

Description

This tool classifies reads a UX-SAD dataset from standard input and, for each line, converts the emotion values into discrete classes.

The tool manages 3 classes, converted to numeric values to ease the training process and to show their natural ordering relation: LOW (0), MEDIUM (1) and HIGH (2).

A new dataset, having as emotion values the number associated with the class, is printed to standard output.

Class Selection

The conversion is done using a simple math condition:

  • If the emotion takes values in the range [0, 100]

    • The class LOW is returned if the emotion value is in the range [0, 33.33)

    • The class MEDIUM is returned if the emotion is in the range [33.33, 66.66)

    • The class HIGH is returned if the emotion is in the range [66.66, 100].

  • If the emotion takes values in the range [-100, 100]

    • The class LOW is returned if the emotion value is in the range [-100, -33.33)

    • The class MEDIUM is returned if the emotion is in the range [-33.33, 33.33)

    • The class HIGH is returned if the emotion is in the range [33.33, 100].

Known Limitations

  • The tool detects the columns to convert by their position (only from the 29th to the 38th column are considered). The columns should, instead, be detected by their name.

  • The tool only supports the conversion using three classes. An option to select the number of classes should be provided.

  • The tool can only read from standard input. An optional argument to read from a file should be provided (to allow usage on systems without I/O redirection).

  • The tool can only write to standard output. An optional argument to write to a file should be provided (to allow usage on systems without I/O redirection).

Copyright

Copyright (C) 2020 Andrea Esposito.
License GPLv3+: GNU GPL version 3 or Later <https://gnu.org/licenses/gpl.html>.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see <https://www.gnu.org/licenses/>.