User Guide

Welcome to the fusionlab-learn User Guide!

This guide provides detailed information on using the library, from core concepts and model architectures to specific features, utilities, data handling, and practical examples.

Explore the sections below to learn how to effectively use fusionlab-learn for your time series forecasting tasks.

Getting Started

• Introduction
  • The Modern Forecasting Challenge
  • Our Philosophy: A Unified and Modular Approach
  • Key Cross-Cutting Features
  • Next Steps
  • References

Core Concepts & Models

• Forecasting Models
  • Base Attentive Model Architecture
  • Hybrid Models
  • Physics-Informed Neural Networks (PINNs)
  • Transformer-Based Models
• Core Model Components
  • Why Understanding Components Matters
  • Architectural Components
  • Input Processing & Embedding Layers
  • Sequence Processing Layers
  • Attention Mechanisms
  • Output & Decoding Layers
  • Loss Function Components
  • Utility Functions
  • Conclusion & Next Steps
• Loss Functions
  • Quantile Loss Functions
  • Anomaly & Combined Loss Functions
  • Loss Function Wrappers/Factories
• Physical Parameter Descriptors
  • General Learnable Parameters (BaseLearnable)
  • Consolidation Coefficient Descriptors (_BaseC)
  • Parameter Resolution Utility (resolve_physical_param)

Data Handling & Utilities

• Datasets
  • Loading Included Datasets
  • Generating Synthetic Datasets
• Data Preparation Workflow
  • Step 1: Imports and Configuration
  • Step 2: Load Raw Data
  • Step 3: Initial Cleaning & Validation
  • Step 4: Feature Engineering
  • Step 5: Feature Selection / Reduction (Optional)
  • Step 6: Define Feature Sets & Scale Numerics
  • Step 7: Reshape into Sequences using reshape_xtft_data
  • Step 8: Train / Validation / Test Split
  • Step 9: Save Processed Data (Optional)
• Utilities for fusionlab-learn
  • Data Manipulation Utilities
  • Time Series Utilities
  • Geospatial & Time Series Data Utilities
  • Spatial Data Utilities
  • Neural Network Utilities
  • PINN Data Utilities
  • Forecast Data Formatting Utilities

Training & Evaluation

• Hyperparameter Tuning
  • Hyperparameter Tuning with HydroTuner
  • HydroTuner: Usage Examples
  • Tuning HALNet with the XTFTTuner
  • TFT Forecast Tuner Guide
  • TFT & XTFT Tuning Examples
  • XTFT Tuning
  • Class-Based Tuner Guide
  • Example: Tuning PIHALNet with PIHALTuner
  • Hyperparameter Tuning (Legacy PiHALTuner)
• Model Evaluation and Visualization
  • Evaluating and Visualizing Forecasts
  • Metrics for Forecasting Evaluation
• Anomaly Detection
  • Anomaly Detection Components
  • Using Anomaly Detection with XTFT

Gallery & Case Studies

• Examples Gallery
  • Forecasting Examples
  • Anomaly Detection Examples
  • Plotting & Visualization Gallery
  • Using Command-Line Tools
  • Tips for Navigating the Gallery:
• Case Histories
  • Nansha Land Subsidence Forecasting (Using XTFT)
  • Zhongshan Land Subsidence Forecasting (Using XTFT)

Exercises

• Hands-On Exercises
  • Exercise: Basic Point Forecasting with Flexible TFT
  • Exercise: Data Preparation Workflow for Case History Data
  • Exercise: Quantile Forecasting with TFT Variants
  • Exercise: Advanced Forecasting with BaseAttentive
  • Exercise: Forecasting with HALNet (All Inputs Required)
  • Exercise: Forecasting with a Pure Transformer
  • Exercise: Forecasting with Stricter TFT (All Inputs Required)
  • Exercise: Advanced Quantile Forecasting with XTFT
  • Exercise: Advanced Forecasting with SuperXTFT
  • Exercise: Hybrid Forecasting with PIHALNet
  • Exercise: Hybrid Forecasting with TransFlowSubsNet
  • Exercise: Physics-Informed Forecasting with PIHALNet & TransFlowSubsNet
  • Exercise: Solving a Forward Problem with PiTGWFlow
  • Exercise: A Basic PIHALNet Forecasting Workflow
  • Exercise: Hyperparameter Tuning with HydroTuner
  • Exercise: Anomaly Detection

Advanced Topics

• Command-Line Tools
  • General TFT Command-Line Interface (CLI)
  • XTFT Probabilistic Prediction Application
  • XTFT Deterministic Prediction Application
• Subsidence PINN Mini GUI Guide
  • Launching the Application
  • User Interface Guide