Tutorial: MC Dropout¶

Notebook: MC_Dropout_Tutorial.ipynb

Purpose¶

Demonstrate predictive uncertainty from dropout-enabled neural networks via Monte Carlo inference.

Data Setup¶

• Beam-like nonlinear regression case
• Structured deflection behavior
• Regions with sparse supervision to stress uncertainty estimates

Core Logic¶

• Train MLP with dropout
• Enable dropout during evaluation
• Aggregate many stochastic passes

Expected Outputs¶

• predictive mean curve
• uncertainty bands that reflect data support and complexity