dno 0.2.1

Dynamic Neural Organism (DNO): A self-evolving, growing, and pruning neural network framework.

DNO: Dynamic Neural Organism 🧬

DNO (Dynamic Neural Organism) is a PyTorch-based framework for creating biological neural networks that grow, think, and evolve at runtime.

Unlike static deep learning models (like Transformers or CNNs), a DNO is an organism that starts small (as a single seed) and physically evolves its architecture based on the problem complexity.

🌟 Why DNO?

	Static Models (Standard AI)	DNO (Dynamic AI)
Architecture	Fixed before training (e.g. 12 layers)	Evolves during training
Adaptability	None (Retraining required)	High (Grows/Shrinks on demand)
Efficiency	Wastes computing on simple tasks	Uses only needed resources
Lifespan	Train once, use forever	Continuous Learning

📦 Installation

pip install dno

🚀 Quick Start

1. The "Hello World" of Life

Create a brain, give it a DNA configuration, and let it think.

import torch
import torch.nn as nn
from dno.core.organism import OrganismManager, BaseEvolvableModule
from dno.core.network import DynamicNetwork
from dno.config import DnoConfig
from dno.utils.dashboard import print_organism_status

# 1. DNA Configuration
# entropy_threshold: Level of "confusion" needed to trigger growth (0.0 - 1.0)
config = DnoConfig(entropy_threshold=0.6, growth_alpha=0.5)

# 2. Birth
manager = OrganismManager()
network = DynamicNetwork(manager, config)

# 3. Seed Layer (The first neuron block)
seed = BaseEvolvableModule(nn.Linear(10, 5))
seed.dynamic_id = "seed_cortex"
network.add_layer(seed)

# 4. Live (Forward Pass)
input_data = torch.randn(1, 10)
output = network(input_data)

print(f"Output: {output.shape}")
print_organism_status(manager)

---

🧠 Core Concepts

1. Neurogenesis (Growth) 📈

When the model is "confused" (High Entropy in outputs) for a sustained period, it undergoes mitosis. It clones its most active layer, adds microscopic noise (mutation) to the clone, and rewires the brain to accommodate the new capacity.

from dno.core.growth import GrowthEngine
import torch.optim as optim

optimizer = optim.SGD(network.parameters(), lr=0.01)
growth_engine = GrowthEngine(network, config)

# ... inside training loop ...
# If model is consistently confused (entropy high):
if growth_engine.check_growth_trigger(entropy_history, current_step):
    print("🧠 Brain is growing...")
    growth_engine.mitosis("seed_cortex", optimizer)

2. Natural Selection (Pruning) ✂️

The SurvivalEngine monitors the Utility Score of every layer. If a layer isn't contributing to information processing (low KL-Divergence between input/output), it is marked for death.

from dno.core.survival import SurvivalEngine

survival = SurvivalEngine(manager, config)

# ... periodically ...
survival.apply_selective_decay(optimizer) # Rot unused weights
survival.garbage_collect(network)         # Remove dead layers

3. Fluid Serialization 💾

Save the entire organism—including its unique topology, weights, and life history—into a single file.

# Save existence
network.save_dno("my_organism.dno")

# Resurrect
new_network = DynamicNetwork.load_dno("my_organism.dno", module_factory=lambda t: nn.Linear(10, 5))

---

🧪 Advanced Usage: Interactive Growth

The DNO can grow based on conversation difficulty. If you feed it simple data, it remains small. If you feed it complex noise (high entropy), it expands.

See examples/interactive_demo.py (or check the repo) for a full simulation where the model adapts to user input complexity in real-time.

🤝 Contributing

DNO is an open-source experiment in Artificial Life.

• Found a bug? Open an issue.
• Have an idea for a new organ? Submit a PR.

📜 License

MIT License. Go build something alive.