Codebase Deep Dive

This section contains a detailed analysis of the Leaspy source code structure, explaining the interaction between Models, Algorithms, and Data.

Contents

• Architecture & Data Flow
  • High-Level Overview
  • Simplified workflow structure
  • Why This Architecture?
• ModelInterface
  • What is an Interface?
  • The Contract: Mandatory Capabilities
  • The Hierarchy: Interface, Base, and Concrete Models
  • Why Do We Need This?
  • Moving to Implementation: BaseModel
• BaseModel
  • The Bridge Between Model, Algorithm, and Data
  • Anatomy of fit(): The Three-Step Orchestration
  • Dimension vs Features: Providing the Output Structure
  • From Abstract to Concrete: The Inheritance Chain
• StatefulModel
  • The “State” Concept
  • Developer Note: When to use which?
  • What comes next?
• McmcSaemCompatibleModel
  • The Optimization Contract
  • Other Responsibilities
  • The Algorithm Loop
• TimeReparametrizedModel
  • Responsibilities
  • Key Methods
  • Key Attributes
  • Next Step
• RiemanianManifoldModel
  • Core Geometric Enforcements
  • Key Variables Introduced
  • Next Step
• LogisticModel
  • The Mathematical Shape
  • Responsibilities
  • Key Attributes & Parameters
  • Key Methods
  • Initialization Logic (Mixin)
  • Next Steps
• LogisticInitializationMixin
  • Why Separate Initialization?
  • Key Methods
• Observation Models
  • The Abstract Base: ObservationModel
  • The Standard: GaussianObservationModel
• The Variables DAG
  • Why does Leaspy need a DAG?
  • What is a DAG, concretely?
  • The Logistic Model’s DAG
  • How VariablesDAG is built
  • Relation to State
• Data & Dataset
  • Data: The User Interface
  • Dataset: The Computational Core
  • Where is Data Handled in the Code?
• Internal Vocabulary
  • Variable Types
  • Algorithmic Concepts
  • Software Architecture Terms