Xpdeep Platform Access#
If you do not have access to the Xpdeep platform and would like to request it, please fill out this form.
π© Contact support: support@xpdeep.com
1. π Getting Started#
1.1 Creating and Managing Users (Admin role)#
- Click the User menu (bottom-left of the screen), then select Settings.
- Open XpViz Users (this page may take a few seconds to load).
- Click Create user, enter the new userβs email address, then confirm with Create.
1.2 Inviting Users to a Project#
- Click the User menu (bottom-left of the screen), then select Settings.
- Open Project Users.
- Under Roles & Permissions, select the target project.
- Click Add Users to invite a user to the selected project.
- To change a userβs role, use the drop-down menu in the Role column.
1.3 Technical Documentation#
Consult our full documentation
β οΈ For users who have previously accessed the Xpdeep documentation, please clear your browser cache or use private browsing to ensure you are viewing the latest version.
Useful sections:
- Installing the Xpdeep package β Installation of the xpdeep package.
- Sota-to-Xpdeep tutorial β Transitioning from a classic PyTorch model to an explainable deep Xpdeep model.
- Case studies β Concrete implementations on tabular and temporal data.
- Explain an Existing Model β Examples of how to explain existing models.
1.3 API Key Generation#
From the XpViz interface:
οΈ β My Account β Settings β API Keys
Use the SaaS API key for your first request.
β οΈ This environment is reserved for tests on medium-sized datasets.
1.4 Limitations#
Given the test environment:
- Training and inference times may not be representative of real performance.
- The number of grouped inferences is limited to 200.
2. π Exploration of Models and Their Explanations (XpViz)#
2.1 Viewing a Trained Model#
Select a model, then click Explain to explore its explanations.
2.2 Generating and Explaining Inferences#
- Go to Model for Inference.
- Select a dataset, then click Edit.
- Choose one or more samples of interest.
- Start generating predictions and the associated explanations.
2.3 Navigation Help#
In the XpViz interface, from: οΈ β My Account β XpViz Tutorial, you have access to:
- A presentation video of several core Xpdeep features using the AdultIncome model.
3. π Advanced Analyses#
3.1 Prescriptive Analysis (How-to)#
Automatically calibrates input data to evolve an initial prediction toward a target prediction.
Example: Identify how to adjust sensors to extend the time before a failure, moving from a forecast of 6 hours to 48 hours.
3.2 Comparative Analysis#
Identifies the factors the model uses similarly or differently when predicting results for two groups of interest.
3.3 Notes#
- These two analyses are currently available for tabular data.
- Prescriptive analysis is also available for temporal data (classification).
- Their generalization to all types of data and tasks is being deployed (you will be informed as soon as they are available).
4. πΌ Access to the XpAct Business Interface#
We also provide access to XpAct, the business-oriented interface, for the Axa and Prostate-Cancer models, from οΈ β My Account β XpAct. The generalization of this interface to other models is in progress.
5. π§© Explainable Models Available in XpViz#
Tabular Data
| Model Name | Data Type | AI Task | Advanced Analysis |
|---|---|---|---|
| Adult Income | Tabular | Classification | How-To, Comparative Analysis |
| Bike | Tabular | Regression | How-To, Comparative Analysis |
| Axa | Tabular | Multi-target regression | How-To, Comparative Analysis |
| Prostate-Cancer | Tabular | Classification | How-To, Comparative Analysis |
Temporal Data
| Model Name | Data Type | AI Task | Advanced Analysis |
|---|---|---|---|
| Human Activity Recognition | Multivariate Time Series | Classification | How-To, Comparative Analysis |
| Air Quality | Multivariate Time Series | Forecasting | Ongoing |
| Scania | Multivariate Time Series | Anomaly Detection, Classification | How-To |
| Virtual Sensors | Multivariate Time Series | Multi-target regression | Ongoing |
Images Data
| Model Name | Data Type | AI Task | Advanced Analysis |
|---|---|---|---|
| Kitti | Images | Object Detection | Ongoing |
| Wolf-Husky | Images | Classification | Ongoing |
| MNIST | Images | Classification | Ongoing |