Skip to content

dashboard.md

Latest commit

 

History

History
180 lines (125 loc) · 4.6 KB

dashboard.md

File metadata and controls

180 lines (125 loc) · 4.6 KB

Interactive Dashboard

ShortKit-ML includes an interactive Gradio web interface for visual exploration and analysis.

Launching the Dashboard

# Ensure dashboard dependencies are installed
pip install "shortcut-detect[dashboard]"

# Launch the dashboard
python app.py

The dashboard opens at http://127.0.0.1:7860

Features

Sample Data

The dashboard comes pre-loaded with CheXpert medical imaging data:

  • 2,000 samples
  • 512-dimensional embeddings
  • Race-based demographic groups
  • Ready for immediate exploration

Custom Data Upload

Upload your own data as a CSV file:

embedding_0,embedding_1,embedding_2,...,task_label,group_label
0.123,0.456,0.789,...,1,group_a
0.234,0.567,0.890,...,0,group_b
...

Requirements:

  • Columns named embedding_0, embedding_1, etc. for embedding dimensions
  • Column named task_label for task labels (optional)
  • Column named group_label for protected attribute labels
  • Optional: attr_<name> columns (e.g., attr_race, attr_gender) for Causal Effect

Detection Methods

Select which detection methods to run:

  • HBAC - Clustering analysis
  • Probe - Classifier-based detection
  • Statistical - Hypothesis testing
  • Geometric - Subspace analysis
  • Causal Effect - Spurious attribute detection (requires attr_* columns)

Visualizations

The dashboard provides interactive visualizations:

  • t-SNE/UMAP plots - Embedding space visualization colored by group
  • Feature importance - Which dimensions show group differences
  • Cluster dendrograms - HBAC hierarchical structure
  • ROC curves - Probe classifier performance

Advanced, optional tools (GradCAM attention overlap, GT mask overlap, SpRAy heatmap clustering, CAV concept testing, and VAE image shortcut detection) are available in the Advanced Analysis tab. You can also optionally include CAV in the main Detection report by uploading a concept bundle.

Report Export

Export your analysis:

  • HTML Report - Interactive report with all visualizations
  • PDF Report - Printable summary document
  • CSV Export - Raw results for further analysis (ZIP file)

Configuration Options

Method Parameters

Parameter Default Description
HBAC max iterations 3 Maximum clustering depth
HBAC min cluster size 0.05 Minimum cluster size as fraction
Probe model Logistic Regression Classifier for probe-based detection
Statistical test Mann-Whitney U Non-parametric test for group differences
Significance level 0.05 Alpha for statistical tests
Causal effect spurious threshold 0.1 Attributes with |effect| below this are flagged
CAV quality threshold 0.7 Minimum concept quality (AUC)
VAE latent dim 10 VAE latent dimension (Advanced Analysis)

Visualization Settings

Setting Default Description
Dimensionality reduction t-SNE Method for 2D projection
Perplexity (t-SNE) 30 t-SNE perplexity parameter
n_neighbors (UMAP) 15 UMAP neighborhood size
Color palette Categorical Color scheme for groups

Keyboard Shortcuts

Shortcut Action
Enter Run analysis
Ctrl+S Download report
Esc Cancel operation

Screenshot

Dashboard Interface

Troubleshooting

Dashboard Won't Start

# Check Gradio is installed
python -c "import gradio; print(gradio.__version__)"

# Reinstall if needed
pip install --upgrade gradio

Port Already in Use

# Use a different port
python app.py --port 7861

Large File Upload Fails

The default upload limit is 100MB. For larger files:

# In app.py, modify the launch command
demo.launch(max_file_size=500)  # 500MB limit

Programmatic Access

You can also run the dashboard programmatically:

from app import create_demo

# Create the Gradio demo
demo = create_demo()

# Launch with custom settings
demo.launch(
    server_name="0.0.0.0",  # Allow external access
    server_port=7860,
    share=True,  # Create public link
)

Docker Deployment

FROM python:3.10-slim

WORKDIR /app
COPY . .

RUN pip install -e ".[dashboard]"

EXPOSE 7860
CMD ["python", "app.py", "--host", "0.0.0.0"]
docker build -t shortcut-detect-dashboard .
docker run -p 7860:7860 shortcut-detect-dashboard

Next Steps