OLED Device Machine Learning Panel

OLED Device Machine Learning Panel Features

Training Data tab

Load, view, edit, and plot OLED device data for training ML models. Data sets are randomly split to a train and test set to build and evaluate the model as specified in the Training set size text box in the Build tab.

Load training data button

Load a CSV file with OLED devices data to train the ML model on. Click to open the OLED device CSV file for training dialog box, where you can navigate to the file. The name of the file you selected is displayed in the text box. This CSV file is copied into the job directory as jobname_input.csv.

The CSV file(s) must contain a specific set of headers to be compatible with the panel.

Each row in the input .csv file is an OLED device with different specifications. Device architectures are defined by layer types, thicknesses, and material composition. The device must be composed of the layer types described in the Using section.

The input data is organized into four column types, where the first three are required to fully define the structural and chemical composition of each layer:

• SMILES Columns—These columns contain information about the layer type, the layer index, SMILES representation of the materials in the layer, and the SMILES index. The column headers follow the format: {Layer_Type}_{Layer_index}_SMILES_{SMILES_index}.

  • Layer_Type—specifies the layer's function (e.g., HIL, ETL). This prefix must be from the list of supported layer types. The emissive layer (EML) has a distinct structure compared to other layers. It is composed of two components: the "Host" material and the "Emitter" dopant material. The same indexing rules apply as with other layers, organizing the "Host" and "Emitter" information into separate columns.

  • Layer_index—indicates the layer's sequential order. This helps organize multiple layers of the same type (e.g., EBL_0, EBL_1). Any number of layers of the same type are allowed.

  • SMILES_index—denotes the index of each component within that layer. Multiple materials within a single layer are represented by sequential SMILES_index values (e.g., Emitter_0_SMILES_0, Emitter_0_SMILES_1).

• Composition Columns—These columns contain the relative composition of the materials in each layer in percentages. The compositions must sum up to 100% for each layer. The column headers follow the format: {Layer_Type}_{Layer_Index}_comp_{Component_Index}

  • The composition columns only need to be specific if the layer is composed of two or more materials. Otherwise, a 100% composition is assumed.

  • For the Emitter layer type, we employ a unique composition specification that follows a different format: Emitter_{Layer_Index}_wt%_{Component_Index}. The associated Host_{Layer_Index} column does NOT require any wt% column. Its composition is determined from the emitter columns: see the example provided.

• Thickness Columns—These columns specify the thickness of each individual layer, in nanometers. The column headers follow the format: {Layer_Type}_{Layer_Index}_thickness. No thickness should be specified for the Emitter layer type.

• Property Columns —These columns contain properties of each OLED device, such as EQE and λ_max. There is no required format for these column names.

See the Using the OLED Device Machine Learning Panel section for more information.

OLED Devices legend

Displays the color of each layer in the device and its corresponding layer type.

OLED Devices input table

Displays the OLED device architectures/structures from the CSV file. Click on the expand icon () to view additional information about the device architecture including the composition and SMILES of each layer. Clicking on SMILES string opens the Component editor, similar to the 2D Sketcher, to edit the component structure. Click OK to save your changes or Cancel to discard any changes. This is helpful to visualize and validate 2D structures instead of manually editing SMILES. To edit any data values, click the Switch to Edit Mode icon (). Click on the Switch to View Mode icon once finished editing (). Editing the data in the panel does not edit the imported CSV file.

Property data from the input CSV file is available to the right of the device architecture information. The size of the two tables can be toggled by dragging the divider between them side to side.

OLED Devices information—The device information is displayed. A stacked layer visualization is displayed, with layer colors corresponding to layer types. The thickness, in nanometers, number of layers, and number of material components in the device are displayed to the right of the stacked layer visualization.

Additional columns—Additional properties specified in the CSV file such as target property or descriptors.

Export Data button

Export data in the OLED Devices input table to a CSV file. Opens the Export OLED Formulations Data dialog box so you can navigate to a location and name the file.

Build tab

Specify parameters for training ML models and the property to be predicted.

Model type options

Specify the model type to use for training:

• Regression—A regression model type is used for the training data set. Numerical values are required for the Target property when using this option.
• Classification—A classification model type is used for the training data set. Binary values are required for the Target property when using this option.

Target property option menu

Specify the properties on which models will be trained and used for prediction. The Target property must be present in the input CSV file. If multiple target properties are selected then individual models are trained sequentially for each.

Additional device descriptors option menu

Select any additional descriptor properties to add to the model training. These properties must be numerical and present in the input CSV file. The property selected in the Target property option menu is not available as a descriptor. By default, none are selected.

Hyperparameter tuning steps option and text box

Set the number of hyperparameter optimization cycles n. Hyperparameters are defined as choices of featurizers and models. In the initial steps, hyperparameters are randomly selected to train the first model. In subsequent steps, the hyperparameters and performance of the previous models are used to select hyperparameters by Bayesian Optimization to maximize the model performance of the next model. A total of n model architectures are explored. The final model uses an ensemble of 3 top-performing models to generate predictions and uncertainties. As a result, a minimum value of 3 is required for this parameter. Increasing values of this setting increases the computation time and model accuracy.

Time limit option and text box

Specify the maximum training time for the model training, in hours. When this amount of time has elapsed, the training is completed for the current model, but no new models are trained after that. The elapsed time can be significantly longer than the limit specified here, if a model takes a long time to train.

Training set size text box

Specify the percentage of the OLED device data which should be used for training the model. The remaining data will be used to test the model.

Pretrained Models option and menu

Select this option to use the predictions from pretrained models as inputs when training a ML model. Choose pretrained models of interest using the menu. Learn more about the models available in the menu in the documentation for the Machine Learning Property Prediction Panel.

Custom DeepAutoQSAR Model option and menu

Select this option to choose and use a model trained from the DeepAutoQSAR Panel and use its predictions as inputs when training a ML model. When this option is selected, the Browse and Delete Selected Models buttons are displayed. Use the Browse button to load DeepAutoQSAR models of interest.

Browse button

Click Browse to open the Select DeepAutoQSAR model dialog box, where you can navigate to the file and click Open. This opens the Enter Model Name dialog box so you can name the DeepAutoQSAR model for use in the panel.

Only available when the Custom DeepAutoQSAR Model option is selected.

Delete Selected Models button

Remove the models selected in the Custom DeepAutoQSAR Model option and menu.

Only available when the Custom DeepAutoQSAR Model option is selected.

Advanced Options button

Set further options for training the ML model. Opens the Training Options dialog box.

Downsample option and text box

Select this option to downsample the data by the specified factor for hyperparameter tuning. This can help speed up training of ML models, particularly for large (> 10,000 structures) training sets.

Out-of-sample splitting option

Select this option to test the model on unique OLED devices not seen in the training set, instead of randomly splitting the data. This option is useful for assessing how well a model might generalize to new devices.

Cross validation splits text box

Specify the number of splits for cross validation.

Random seed for training/test set splitting text box

Select this option to specify a random seed to be used for splitting the training and test set.

Correlation threshold text box

Specify the threshold for removing highly correlated features.

Performance tab

Analyze the quality of a trained model. A summary of the statistics of the model is presented alongside a plot.

Property option menu and Load Model button

Load a pre-trained ML model or a custom ML model generated using the OLED Device Machine Learning panel. For custom models, click the Load Model button to open the Select OLED Models dialog box, where you can load and manage the ML models. Select the models to analyze the performance of from the option menu after loading.

The pre-trained ML model menu items include Current Efficiency, Color-Index X-coordinate, Color Index Y-coordinate, Electroluminescence Maximum Peak Position, External Quantum Efficiency, Electroluminescence Bandwidth, and Power Efficiency.

Training parameters summary

This section displays the parameters used to train the selected ML model.

Target text

Displays the target property when generating the ML model. Noneditable.

Plot options

Select an option to modify the plot appearance in the tabs below. The options include:

• Same x and y axis—Select this option to enforce the x and y axis to have the same scale and range.

• Plot XY—Select this option to plot a dashed line for x=y.

• Show marginals—Select this option to display individual distributions of the x-axis and y-axis.

• Show stats—Select this option to display the relevant statistics for the model on the plot.

Parity plot area

This area displays a scatter plot of the predicted versus observed target property values. Only present for Regression models.

ROC tab

This area displays a receiver operating characteristic (ROC) curve of the true positive rate versus false positive rate. An ideal model would have a ROC plot shifted upper left with a True Positive Rate of 1, False Positive Rate of 0, and Area Under the Curve of 1. Only present for Classification models.

Confusion Matrix tab

This area displays a confusion matrix for the train and test sets. Only present for Classification models.

Model summary table

For Regression models, this table lists the R-squared value (R2) and the root-mean-square error (RMSE) for the Train and Test sets.

For Classification models, this table lists the area under the ROC curve (Roc Auc), F1 Sensitivity, F1 Specificity, F1 Precision, Accuracy, and F1 Score for the Train and Test sets.

Predict tab

Choose devices to make predictions for and the model to apply. When the job finishes, the results are saved to a CSV file in the job directory with the name jobname_predict.csv.

Property option menu and Load Model button

Load a pre-trained ML model or a custom ML model generated using the OLED Device Machine Learning panel. For custom models, click the Load Model button to open the Select OLED Models dialog box, where you can load and manage the ML models. Select the models to use for the prediction task from the option menu after loading.

The pre-trained ML model menu items include Current Efficiency, Color-Index X-coordinate, Color Index Y-coordinate, Electroluminescence Maximum Peak Position, External Quantum Efficiency, Electroluminescence Bandwidth, and Power Efficiency.

Predict Data option menu and Load button

Choose to load a test set for prediction or the results of a prediction calculation for review.

Prediction Input—Load a CSV file with OLED device data for which we want to predict the target property. Click Load to open the OLED device CSV file for predictions dialog box, where you can navigate to the file. The name of the file you selected is displayed in the text box. This CSV file will be copied into the job directory as jobname_input.csv. See the Using the OLED Device Machine Learning Panel section for a description of the required format for the CSV file.

Prediction Output—Load a CSV file generated from a job run from the Predict tab of the OLED Device Machine Learning panel. Click Load to open the OLED device CSV file with predictions output dialog box, where you can navigate to the file. The name of the file you selected is displayed in the text box. The file must have an extension of jobname_predict.csv.

Prediction parameters summary

This section displays the parameters pertinent to using the loaded ML model for property prediction.

Target text

Displays the target property for the ML model. Noneditable.

Descriptors text

Lists the additional descriptors used in generating the ML model. These descriptors must be available in the prediction data set if they were used to train the ML models. Noneditable.

OLED Devices legend

Displays the color of each layer in the device and its corresponding layer type.

OLED Formulations results table

Displays the OLED device architectures/structures from the jobname_input.csv or jobname_prediction.csv file. Click on the expand icon () to view additional information about the device architecture including the composition and SMILES of each layer.

Property data from the input CSV file is available to the right of the device architecture information. The size of the two tables can be toggled by dragging the divider between them side to side.

OLED Devices information—The device information is displayed. A stacked layer visualization is displayed, with layer colors corresponding to layer types. The thickness, in nanometers, number of layers, and number of material components in the device are displayed to the right of the stacked layer visualization.

Additional columns—Additional properties specified in the CSV file such as target property or descriptors. For results data, the prediction values and uncertainties are appended to the list.

Job toolbar

Manage job submission and settings. See Job Toolbar for a description of this toolbar.

The Job Settings button opens the OLED Device Machine Learning - Job Settings Dialog Box, where you can make settings for running the job.

Status bar

Use the Reset button to reset the panel to its default settings and clear any data from the panel. If the panel has a Job toolbar, you can also reset the panel from the Settings button menu.

If you can submit a job from the panel, the status bar displays information about the current job settings and status for the panel. The settings include the job name, task name and task settings (if any), number of subjobs (if any) and the host name and job incorporation setting. The job status can include messages about job start, job completion and incorporation.

The status bar also contains the Help button , which opens an option menu with choices to open the help topic for the panel (Documentation), launch Maestro Assistant, or if available, choose from an option menu of Tutorials. If the panel is used by one or more tutorials, hover over the Tutorials option to display a list of tutorials. Choosing a tutorial opens the tutorial topic.

Related Topics

• Optoelectronic Device Designer Panel
• Machine Learning Property Prediction Panel
• ML Model Manager Panel
• OLED Device Machine Learning - Job Settings Dialog Box