Error Analysis

https://erroranalysis.ai
Admin
Error Analysis is a Responsible AI toolkit that enables machine learning practitioners to identify...
Tokens:1,906
Snippets:16
Trust Score:7.5
Update:1 month ago
Show doc for...
Tokens:
Raw
### ExplanationDashboard

Source: https://erroranalysis.ai/raiwidgets.html

Initializes the Explanation dashboard to visualize model explanations.

```APIDOC
## class ExplanationDashboard(explanation, model=None, dataset=None, true_y=None, classes=None, features=None, public_ip=None, port=None, locale=None)

### Description
Wraps the explanation dashboard component to visualize model explanations.

### Parameters
- **explanation** (ExplanationMixin) - Required - Object representing an explanation.
- **model** (object) - Optional - Model object with predict() or predict_proba() methods.
- **dataset** (numpy.ndarray/list) - Optional - Feature matrix.
- **true_y** (numpy.ndarray/list) - Optional - True labels.
- **classes** (numpy.ndarray/list) - Optional - Class names.
- **features** (numpy.ndarray/list) - Optional - Feature names.
- **public_ip** (str) - Optional - External IP for remote VM.
- **port** (int) - Optional - Port for local service.
- **locale** (str) - Optional - Language code (default 'en').
```

--------------------------------

### Run Error Analysis with Model and Explanation

Source: https://erroranalysis.ai/raiwidgets.html

Initialize the dashboard using a model object and a pre-computed explanation.

```python
>>> from raiwidgets import ErrorAnalysisDashboard
>>> ErrorAnalysisDashboard(global_explanation, model,
...                        dataset=X_test, true_y=y_test)
```

--------------------------------

### Run ErrorAnalysisDashboard

Source: https://erroranalysis.ai/raiwidgets.html

Initializes the ErrorAnalysisDashboard with both sample and full datasets for model error evaluation.

```python
>>> from raiwidgets import ErrorAnalysisDashboard
>>> ErrorAnalysisDashboard(sample_dataset=X_test_sample,
...                        dataset=X_test,
...                        features=features,
...                        true_y=y_test_sample,
...                        true_y_dataset=y_test,
...                        pred_y=X_test_sample_pred_y,
...                        pred_y_dataset=X_test_pred_y)
```

--------------------------------

### FairnessDashboard

Source: https://erroranalysis.ai/raiwidgets.html

Initializes the Fairness dashboard to evaluate model fairness metrics.

```APIDOC
## class FairnessDashboard(sensitive_features, y_true, y_pred, locale=None, public_ip=None, port=None, fairness_metric_module=None, fairness_metric_mapping=None)

### Description
Wraps the fairness dashboard component to evaluate model predictions against sensitive features.

### Parameters
- **sensitive_features** (pandas.Series/DataFrame/list/dict) - Required - Sensitive features for fairness analysis.
- **y_true** (numpy.ndarray/list) - Required - True labels.
- **y_pred** (pandas.Series/DataFrame/list/dict) - Required - Model predictions.
- **locale** (str) - Optional - Language code.
- **public_ip** (str) - Optional - External IP for remote VM.
- **port** (int) - Optional - Port for local service.
```

--------------------------------

### Run Error Analysis with Predictions and True Labels

Source: https://erroranalysis.ai/raiwidgets.html

Initialize the dashboard using only prediction results and ground truth labels.

```python
>>> predictions = model.predict(X_test)
>>> from raiwidgets import ErrorAnalysisDashboard
>>> ErrorAnalysisDashboard(dataset=X_test, true_y=y_test,
...                        features=features, pred_y=predictions)
```

--------------------------------

### ErrorAnalysisDashboard

Source: https://erroranalysis.ai/raiwidgets.html

The ErrorAnalysisDashboard class is used to initialize and launch an interactive error analysis dashboard for model evaluation.

```APIDOC
## class raiwidgets.ErrorAnalysisDashboard

### Description
Initializes the ErrorAnalysisDashboard to visualize model performance, error patterns, and explanations. It supports both classification and regression tasks.

### Parameters
- **explanation** (ExplanationMixin) - Optional - An object representing an explanation.
- **model** (object) - Optional - The model object, expected to have predict_proba() for classification or predict() for regression.
- **dataset** (pd.DataFrame/numpy.ndarray/list) - Optional - The feature matrix used for analysis.
- **true_y** (numpy.ndarray/list) - Optional - The true labels for the dataset.
- **classes** (numpy.ndarray/list) - Optional - Class names for classification tasks.
- **features** (numpy.ndarray/list) - Optional - Feature names.
- **port** (int) - Optional - Port for the locally hosted service.
- **public_ip** (str) - Optional - External public IP address if running on a remote VM.
- **categorical_features** (list[str]) - Optional - Names of categorical features.
- **true_y_dataset** (numpy.ndarray/list) - Optional - True labels for the provided dataset.
- **pred_y** (numpy.ndarray/list) - Optional - Predicted values.
- **pred_y_dataset** (numpy.ndarray/list/pandas.Series) - Optional - Predicted labels for the provided dataset.
- **model_task** (str) - Optional - Specifies if the model is for classification or regression.
- **metric** (str) - Optional - Metric to evaluate at tree nodes or heatmap grids.
- **max_depth** (int) - Optional - Maximum depth of the surrogate tree.
- **num_leaves** (int) - Optional - Number of leaves in the surrogate tree.
- **min_child_samples** (int) - Optional - Minimum samples required for a leaf.
- **sample_dataset** (pd.DataFrame/numpy.ndarray/list) - Optional - A subset of the dataset for performance optimization.
- **locale** (str) - Optional - Language for the dashboard interface.

### Usage Example
```python
from raiwidgets import ErrorAnalysisDashboard
ErrorAnalysisDashboard(dataset=X_test, true_y=y_test, features=features, pred_y=predictions)
```
```

--------------------------------

### raiwidgets.ResponsibleAIDashboard

Source: https://erroranalysis.ai/raiwidgets.html

The ResponsibleAIDashboard class is the primary interface for initializing and hosting the Responsible AI dashboard component.

```APIDOC
## class raiwidgets.ResponsibleAIDashboard

### Description
The ResponsibleAIDashboard class wraps the dashboard component to visualize RAIInsights analysis.

### Parameters
- **analysis** (RAIInsights) - Required - An object that represents an RAIInsights analysis.
- **public_ip** (str) - Optional - The external public IP address of the VM if running on a remote VM.
- **port** (int) - Optional - The port to use on the locally hosted service.
- **locale** (str) - Optional - The language in which to load the dashboard (default is "en").
- **cohort_list** (List[Cohort]) - Optional - List of cohorts defined by the user.
- **is_private_link** (bool) - Optional - Indicates if the dashboard environment is a private link AML workspace.
```

--------------------------------

### ErrorAnalysisDashboard

Source: https://erroranalysis.ai/raiwidgets.html

Initializes the Error Analysis dashboard to visualize model errors on large and downsampled datasets.

```APIDOC
## class ErrorAnalysisDashboard(sample_dataset, dataset, features, true_y, true_y_dataset, pred_y, pred_y_dataset)

### Description
Initializes the ErrorAnalysisDashboard component for analyzing model performance and errors.

### Parameters
- **sample_dataset** (numpy.ndarray/list) - Required - Downsampled dataset for UI.
- **dataset** (numpy.ndarray/list) - Required - Full dataset.
- **features** (numpy.ndarray/list) - Required - Feature names.
- **true_y** (numpy.ndarray/list) - Required - True labels for sample dataset.
- **true_y_dataset** (numpy.ndarray/list) - Required - True labels for full dataset.
- **pred_y** (numpy.ndarray/list) - Required - Predictions for sample dataset.
- **pred_y_dataset** (numpy.ndarray/list) - Required - Predictions for full dataset.
```
Tokens:
Raw