<tool id="ml_visualization_ex" name="Machine Learning Visualization Extension" version="@VERSION@">
    <description>includes several types of plotting for machine learning</description>
    <macros>
        <import>main_macros.xml</import>
        <import>keras_macros.xml</import>
    </macros>
    <expand macro="python_requirements"/>
    <expand macro="macro_stdio"/>
    <version_command>echo "@VERSION@"</version_command>
    <command>
        <![CDATA[
        python '$__tool_directory__/ml_visualization_ex.py' 
            --inputs '$inputs'
            #if $plotting_selection.plot_type in ['learning_curve', 'rfecv_gridscores']
            --infile1 '$plotting_selection.infile1'
            #elif $plotting_selection.plot_type in ['pr_curve', 'roc_curve']
            --infile1 '$plotting_selection.infile1'
            --infile2 '$plotting_selection.infile2'
            #elif $plotting_selection.plot_type == 'feature_importances'
            --estimator '$plotting_selection.infile_estimator'
            --infile1 '$plotting_selection.infile1'
            #elif $plotting_selection.plot_type == 'keras_plot_model'
            --model_config '$plotting_selection.infile_model_config'
            #end if
        ]]>
    </command>
    <configfiles>
        <inputs name="inputs" />
    </configfiles>
    <inputs>
        <conditional name="plotting_selection">
            <param name="plot_type" type="select" label="Select a plotting type">
                <option value="learning_curve" selected="true">Learning curve</option>
                <option value="pr_curve">2-class / multpi-label Precison Recall curve</option>
                <option value="roc_curve">2-class / multi-label Receiver Operating Characteristic (ROC) curve</option>
                <option value="rfecv_gridscores">Number of features vs. Recursive Feature Elimination gridscores with corss-validation</option>
                <option value="feature_importances">Feature Importances plot</option>
                <option value="keras_plot_model">keras plot model - plot configuration of a neural network model</option>
            </param>
            <when value="learning_curve">
                <param name="infile1" type="data" format="tabular" label="Select the dataset containing values for plotting learning curve." help="This dataset should be the output of tool model_validation->learning_curve."/>
                <param name="plot_std_err" type="boolean" truevalue="booltrue" falsevalue="boolfalse" optional="true" checked="false" label="Whether to plot standard error bar?"/>
                <param name="title" type="text" value="" optional="true" label="Plot title" help="Optional. If change is desired."/>
                <param name="plot_format" type="select" label="The output format and library">
                    <option value="html" selected="true">An interactive html by plotly</option>
                    <!--<option value="png">PNG image by matplotlib</option> TODO-->
                </param>
            </when>
            <when value="pr_curve">
                <param name="infile1" type="data" format="tabular" label="Select the dataset containing true labels." help="y_true. Headered. For 2-class, single column contains both class labels (e.g. True and False). For multi-label, each column, hot-encoded, corresponds to one label."/>
                <param name="infile2" type="data" format="tabular" label="Select the dataset containing predicted probabilities." help="y_preds. Headered. For 2-class, sinle column or the first column contains scores for the positive label. For multi-label, each column corresponds to one label."/>
                <param name="pos_label" type="text" value="" optional="true" label="pos_label" help="The label of positive class. If not specified, it will be 1 by default."/>
                <param name="title" type="text" value="" optional="true" label="Plot title" help="Optional. If change is desired."/>
                <param name="report_minimum_n_positives" type="integer" value="" optional="true" label="Report minimum number of positives" help="For mulitple label binary classifications, whose number of true postives is less than the threhold will be ignored."/>
                <param name="plot_format" type="select" label="The output format and library">
                    <option value="plotly_html" selected="true">An interactive html by plotly</option>
                    <option value="matplotlib_svg">SVG image by matplotlib</option>
                </param>
            </when>
            <when value="roc_curve">
                <param name="infile1" type="data" format="tabular" label="Select the dataset containing true labels." help="y_true. Headered. For 2-class, single column contains both class labels (e.g. True and False). For multi-label, each column, hot-encoded, corresponds to one label."/>
                <param name="infile2" type="data" format="tabular" label="Select the dataset containing predicted probabilities." help="y_preds. Headered. For 2-class, sinle column or the first column contains scores for the positive label. For multi-label, each column corresponds to one label."/>
                <param name="pos_label" type="text" value="" optional="true" label="pos_label" help="The label of positive class. If not specified, it will be 1 by default."/>
                <param name="drop_intermediate" type="boolean" truevalue="booltrue" falsevalue="boolfalse" optional="true" checked="true" label="drop_intermediate" help="Whether to drop some suboptimal thresholds which would not appear on a plotted ROC curve."/>
                <param name="title" type="text" value="" optional="true" label="Plot title" help="Optional. If change is desired."/>
                <param name="report_minimum_n_positives" type="integer" value="" optional="true" label="Report minimum number of positives" help="For mulitple label binary classifications, whose number of true postives is less than the threhold will be ignored."/>
                <param name="plot_format" type="select" label="The output format and library">
                    <option value="plotly_html" selected="true">An interactive html by plotly</option>
                    <option value="matplotlib_svg">SVG image by matplotlib</option>
                </param>
            </when>
            <when value="rfecv_gridscores">
                <param name="infile1" type="data" format="tabular" label="Select the dataset containing grid_scores from a fitted RFECV object." help="Headered. Single column. Could be Output from `estimator_attributes->grid_scores_`."/>
                <param name="steps" type="text" value="" optional="true" label="Step IDs" help="List, containing hover labels for each grid_score. For example: `list(range(10)) + [15, 20]`."/>
                <param name="title" type="text" value="" optional="true" label="Plot title" help="Optional. If change is desired."/>
                <param name="plot_format" type="select" label="The output format and library">
                    <option value="html" selected="true">An interactive html by plotly</option>
                    <!--<option value="png">PNG image by matplotlib</option> TODO-->
                </param>
            </when>
            <when value="feature_importances">
                <param name="infile_estimator" type="data" format="zip" label="Select the dataset containing fitted estimator/pipeline" />
                <param name="infile1" type="data" format="tabular" label="Select the dataset containing feature names" help="Make sure the headers (first row) are feature names."/>
                <conditional name="column_selector_options">
                    <expand macro="samples_column_selector_options" multiple="true"/>
                </conditional>
                <param name="threshold" type="float" value="" optional="true" min="0." label="Threshold value" help="Features with importance below the threshold value will be ignored."/>
                <param name="title" type="text" value="" optional="true" label="Custom plot title" help="Optional."/>
                <param name="plot_format" type="select" label="The output format and library">
                    <option value="html" selected="true">An interactive html by plotly</option>
                    <!--<option value="png">PNG image by matplotlib</option> TODO-->
                </param>
            </when>
            <when value="keras_plot_model">
                <param name="infile_model_config" type="data" format="json" label="Select the JSON dataset containing configuration for a neural network model"/>
                <param name="title" type="hidden" value="" optional="true" label="Plot title" help="Optional. If change is desired."/>
                <param name="plot_format" type="hidden" value="png" label="The output format and library"/>
            </when>
        </conditional>
    </inputs>
    <outputs>
        <data name="output" format="html" from_work_dir="output" label="${plotting_selection.plot_type} plot on ${on_string}">
            <change_format>
                <when input="plotting_selection.plot_format" value="png" format="png"/>
            </change_format>
        </data>
    </outputs>
    <tests>
        <test>
            <param name="plot_type" value="roc_curve"/>
            <param name="infile1" value="y_true.tabular" ftype="tabular"/>
            <param name="infile2" value="y_score.tabular" ftype="tabular"/>
            <output name="output" file="ml_vis04.html" compare="sim_size"/>
        </test>
        <test>
            <param name="plot_type" value="feature_importances"/>
            <param name="infile_estimator" value="best_estimator_.zip" ftype="zip"/>
            <param name="infile1" value="regression_X.tabular" ftype="tabular"/>
            <param name="selected_column_selector_option" value="all_columns"/>
            <output name="output" file="ml_vis01.html" compare="sim_size"/>
        </test>
        <test>
            <param name="plot_type" value="learning_curve"/>
            <param name="infile1" value="mv_result03.tabular" ftype="tabular"/>
            <output name="output" file="ml_vis02.html" compare="sim_size"/>
        </test>
        <test>
            <param name="plot_type" value="pr_curve"/>
            <param name="infile1" value="y_true.tabular" ftype="tabular"/>
            <param name="infile2" value="y_score.tabular" ftype="tabular"/>
            <output name="output" file="ml_vis03.html" compare="sim_size"/>
        </test>
        <test>
            <param name="plot_type" value="rfecv_gridscores"/>
            <param name="infile1" value="grid_scores_.tabular" ftype="tabular"/>
            <output name="output" file="ml_vis05.html" compare="sim_size"/>
        </test>
        <test>
            <param name="plot_type" value="keras_plot_model"/>
            <param name="infile_model_config" value="deepsear_1feature.json" ftype="json"/>
            <output name="output" file="ml_vis05.png" compare="sim_size" delta="20000"/>
        </test>
    </tests>
    <help>
        <![CDATA[
**What it does**
This tool ouputs serveral machine learning visualization plots using Plotly, including 'feature_importances', 'learning curve', 'precison recall curve', 'roc_curve',  and 'number of featues vs. rfecv gridscores'. This tool also ouputs configuration diagram for a deep learning model using the Keras model JSON file as input.


**Feature importances**

.. image:: https://raw.githubusercontent.com/goeckslab/Galaxy-ML/master/galaxy_ml/tools/images/feature_importances.png
    :width: 400
    :alt: feature importances


**Learning curve**

.. image:: https://raw.githubusercontent.com/goeckslab/Galaxy-ML/master/galaxy_ml/tools/images/learning_curve.png
    :width: 400
    :alt: learning curve


**Precison recall curve**

.. image:: https://raw.githubusercontent.com/goeckslab/Galaxy-ML/master/galaxy_ml/tools/images/pr_curve.png
    :width: 400
    :alt: precison recall curve


**Receiver operating characteristic curve**

.. image:: https://raw.githubusercontent.com/goeckslab/Galaxy-ML/master/galaxy_ml/tools/images/roc_curve.png
    :width: 400
    :alt: Receiver operating characteristic curve


**Number of featues vs. rfecv gridscores**

.. image:: https://raw.githubusercontent.com/goeckslab/Galaxy-ML/master/galaxy_ml/tools/images/rfecv_gridscore.png
    :width: 400
    :alt: Number of featues vs. rfecv gridscores


**Deep learning model configuration**

.. image:: https://raw.githubusercontent.com/goeckslab/Galaxy-ML/master/galaxy_ml/tools/images/deepsea.png
    :width: 400
    :alt: deapsea model configuration
        ]]>
    </help>
    <expand macro="sklearn_citation">
        <citation type="bibtex">
            @online{plotly,
            author = {Plotly Technologies Inc.},
            title = {Collaborative data science},
            publisher = {Plotly Technologies Inc.},
            address = {Montreal, QC},
            year = {2015},
            url = {https://plot.ly}}
        </citation>
    </expand>
</tool>