EasyHybrid.jl

Appearance

25

Tutorials Overview ​

Welcome to the EasyHybrid.jl tutorials! This section provides comprehensive guides and examples to help you get started with hybrid modeling—combining process-based models with neural networks.

Getting Started ​

If you're new to EasyHybrid.jl, we recommend starting with the Getting Started guide, which walks through building your first hybrid model step by step.

Tutorial Topics ​

Core Tutorials ​

Exponential Respiration Model ​

Learn to build a hybrid model for soil respiration using an exponential temperature relationship. This tutorial demonstrates:

  • Creating synthetic data with exponential temperature response

  • Defining a process-based model (Expo_resp_model)

  • Configuring model parameters and constructing hybrid models

  • Training and evaluating your model

Best for: Understanding the fundamentals of hybrid modeling with a simple, well-documented example.

Sequence Hybrid Models (LSTM & Transformer) ​

Explore advanced neural network architectures by building a hybrid model with LSTM (Long Short-Term Memory) networks. This tutorial covers:

  • Using LSTM networks for sequence modeling

  • Configuring feedforward vs. recurrent architectures

  • Working with sequence data (input/output windows, lead times)

  • Comparing LSTM and standard neural network performance

Best for: Working with time series data that requires memory of past states.

Model Evaluation and Optimization ​

Cross-Validation ​

Implement k-fold cross-validation to robustly evaluate your hybrid models. Learn how to:

  • Create and manage data folds

  • Train models across multiple validation splits

  • Parallelize cross-validation training

  • Organize results from multiple model runs

Best for: Ensuring your model generalizes well and avoiding overfitting.

Hyperparameter Tuning ​

Optimize your model's performance using Hyperopt.jl for automated hyperparameter search. This tutorial shows:

  • Setting up hyperparameter search spaces

  • Using the tune function for optimization

  • Comparing model performance before and after tuning

  • Best practices for hyperparameter selection

Best for: Finding optimal model configurations and improving performance.

Advanced Topics ​

Losses and LoggingLoss ​

Deep dive into the loss function system in EasyHybrid.jl. Learn about:

  • Predefined loss functions (MSE, MAE, NSE)

  • Creating custom loss functions

  • Handling missing values and uncertainty

  • Using LoggingLoss for training and evaluation

  • Passing additional arguments and keyword arguments to losses

Best for: Customizing model training and implementing domain-specific loss functions.

Slurm Jobs ​

Run EasyHybrid.jl models on HPC clusters using Slurm job scheduling. This tutorial provides:

  • Example Slurm batch scripts

  • Configuring Julia for cluster environments

  • Running array jobs for parallel experiments

  • Resource allocation best practices

Best for: Scaling up training to high-performance computing environments.

Choosing the Right Tutorial ​

Next Steps ​

After completing the tutorials, explore the Research section to see real-world applications, or dive into the API Reference for detailed function documentation.