adaptive-dynamics 0.1.0

Adaptive Dynamics Toolkit: Adaptive π geometry, ARP optimizers, simulations, and compression.

Adaptive Dynamics Toolkit (ADT)

A unified framework for adaptive computing paradigms, including adaptive π geometry, ARP optimization, physics simulations, and compression algorithms.

Installation

pip install adaptive-dynamics

For development or to include optional dependencies:

pip install "adaptive-dynamics[torch,sympy,dev]"
# or with uv
uv venv && uv pip install -e ".[dev,docs,torch,sympy]"

Quick Examples

Curved Geometry with Adaptive π (πₐ)

from adaptive_dynamics.pi.geometry import AdaptivePi

# Create an instance with gentle positive curvature
pi = AdaptivePi(curvature_fn=lambda x, y: 1e-3)

# Calculate circumference in curved space
circumference = pi.circle_circumference(1.0)
print(f"Circumference of unit circle: {circumference:.6f}")
# Output: Circumference of unit circle: 3.144159

Neural Network Training with ARP Optimizer

import torch
import torch.nn as nn
from adaptive_dynamics.arp.optimizers import ARP

# Define a simple model
model = nn.Sequential(nn.Flatten(), nn.Linear(28*28, 10))

# Use ARP optimizer
opt = ARP(model.parameters(), lr=3e-3, alpha=0.01, mu=0.001)

# Training loop (example)
# X, y = ... load a batch ...
loss_fn = nn.CrossEntropyLoss()
loss = loss_fn(model(X), y)
loss.backward()
opt.step()
opt.zero_grad()

Examples

• πₐ Curved Circles
• ARP Optimizer on MNIST

Documentation

Full documentation is available at https://RDM3DC.github.io/adaptive-dynamics-toolkit

• Getting Started Guide
• API Reference
• Tutorials

Features

• Adaptive π Geometry: Curved space mathematics and Gauss-Bonnet inspired algorithms
• ARP Optimization: Resistance-conductance model for neural network optimization
• Physics Simulations: Gravity, beams, and ringdown simulations with adaptive precision
• Compression Tools: Adaptive compression for text, curves, and tensors
• TSP Solvers: Tools for 3D printing toolpath optimization

Pro Features

ADT Pro extends the toolkit with advanced features for enterprise and research:

• Advanced CUDA acceleration
• Premium simulation capabilities
• Enterprise-grade dashboards
• Specialized slicer algorithms

Contact us at [e-mail] for licensing information.

Contributing

Contributions are welcome! See CONTRIBUTING.md for details.

License

• Adaptive Dynamics Toolkit (Community Edition): MIT License
• ADT Pro: Commercial license (see pro/README_PRO.md)

Support & Services

• Integration & Research Consulting
• Training & Workshops
• GitHub Sponsors