Skip to content

optframe

Welcome to the Oracle Framework documentation! This framework provides a flexible and modular approach to distributed optimization experiments.

Overview

The Oracle Framework is designed for researchers and practitioners working on distributed optimization algorithms. It provides:

  • Modular Architecture: Separate components for problems, oracles, optimizers, and experiments
  • Flexible Oracle System: Support for deterministic and stochastic oracles
  • Multiple Problem Types: Neural networks, OLS regression, and custom problems
  • Experiment Tracking: Built-in MLflow integration for experiment management

Quick Start

# Install dependencies
uv sync --extra docs

# View documentation locally
uv run mkdocs serve

Then visit http://localhost:8000

Key Components

Problems

Define optimization problems with standardized interfaces. See Problems API.

Oracles

Provide function evaluations and gradients. See Oracles API.

Optimizers

Implement optimization algorithms like GD and AGD. See Optimizers API.

Experiments

Run and track optimization experiments. See Experiments API.

Architecture

For a detailed overview of the system architecture and class relationships, see the Architecture Overview and Class Diagrams.

Guides