(Network Builder)
(v2.0)
A New Neural Growth & Computation Blueprint By Qubic Open Science David Vivancos & José Sanchez
Continuous-time • Trinary-state • Multi-timescale • Neuromodulated • Structurally plastic
Full theoretical foundation: Neuraxon v2.0 Paper (ResearchGate) preprint soon at IEEE Xplore
Core implementation: neuraxon2.py (pure Python, no external dependencies, backward-compatible with v1).
CuNxon Cuda Kernels & Library for Multi Neuraxon 2.0: (https://github.com/DavidVivancos/Neuraxon/cuNxon)
Interactive 3D Network Builder Demo : Hugging Face Space
Multi Neuraxon 2.0 Multi-Sphere 3D Network Builder Demo: : (https://huggingface.co/spaces/DavidVivancos/MultiNeuraxon2)
(Update March 31st 2026: Include brain preset for the abstraction of the fruit fly connectome: Drosophila melanogaster.)
Core implementation: MultiNeuraxon2.py
Neuraxon Game of Life 5.1 v196 Research Version Mulit Nxon 2.0 (Based on Neuraxon 2.0) + First g factor introduced + Evolutionary NAS: (https://github.com/DavidVivancos/Neuraxon/tree/main/GameOfLife/5)
Neuraxon Live -> Nxon GoL Server + Client V 1.064 v1084 (oct 5th 2026) Including NxonKaleido Brain viz : (https://github.com/DavidVivancos/Neuraxon/tree/main/GameOfLife/NxonLive)
Play online at: https://nxon.online/
Neuraxon Game of Life 5.0 Lite Demo (Based on Multi Neuraxon 2.0) + First Proto-Language/Audition : Neuraxon Game of Life Hugging Face Space
Neuraxon is a bio-inspired computational unit that goes far beyond the classic perceptron. It uses trinary logic (+1 excitatory, 0 neutral, -1 inhibitory), operates in continuous time (inputs flow as constant streams), and performs computation at both neuron and synapse levels. Synapses have three dynamic weights (w_fast, w_slow, w_meta) and can form, collapse, or reconnect; hidden neurons can even die. Spontaneous activity, autoreceptors, homeostatic plasticity, and four neuromodulators (DA, 5-HT, ACh, NA) with nine receptor subtypes make the network intrinsically alive and adaptive.
v2.0 introduces a unified 4-step pipeline executed every simulation step:
- Time Warping (ChronoPlasticity) – adaptive memory horizon per synapse
- Dynamic Decay (DSN) – input-conditioned decay via causal convolution
- CTSN Complemented State – learnable complement term prevents information loss
- AGMP (Astrocyte-Gated Multi-timescale Plasticity) – eligibility × modulator × astrocyte gate
Additional breakthroughs:
- MSTH (Multi-Scale Temporal Homeostasis) – 4 coordinated regulatory loops (ultrafast → slow)
- Nonlinear dendritic branch integration with supralinear gamma
- Watts-Strogatz small-world topology (ring + rewiring)
- Full receptor subtype system (tonic/phasic, nonlinear activation)
- Oscillator bank with cross-frequency coupling (PAC)
- Aigarth Intelligent Tissue hybridization for evolutionary mutation/selection
Traditional ANNs and even most SNNs suffer from discrete time steps, binary/spiking simplification, static topologies, catastrophic forgetting, and catastrophic rigidity. Neuraxon 2.0 solves these at the architectural level:
- Continuous real-time learning – no separate training/inference phases; adapts instantly to streaming data.
- Temporal richness – timing, duration, frequency, and sequence of trinary states encode information.
- Biological plausibility – dendrites, silent synapses, neuromodulation, homeostasis, astrocyte-like gating, spontaneous activity.
- Energy efficiency & robustness – sparse trinary activity + multi-timescale decay + homeostasis prevent saturation/plasticity loss.
- Evolvability – Aigarth hybrid enables population-level evolution of network structure and parameters.
Applications that become possible:
- Embodied / robotic control with proprioception and continuous sensory streams
- Real-time temporal pattern recognition (finance, neuroscience, robotics)
- Cognitive modeling of consciousness, attention, sleep/wake cycles
- Continual lifelong learning without forgetting
- Energy-efficient edge AI and neuromorphic hardware targets
- Artificial life and open-ended evolution
- Pathways toward AGI that respects biological constraints
v2.0 is not an incremental update — it is a new blueprint for neural growth and computation.
from neuraxon2 import (
NetworkParameters,
NeuraxonNetwork,
NeuraxonApplication,
NeuraxonAigarthHybrid
)
# 1. Create network with biologically-plausible defaults
params = NetworkParameters(
num_input_neurons=5,
num_hidden_neurons=30,
num_output_neurons=5,
# v2.0 new parameters
dsn_enabled=True,
ctsn_enabled=True,
agmp_enabled=True,
chrono_enabled=True,
msth_ultrafast_tau=5.0,
ws_k=8,
ws_beta=0.35
)
network = NeuraxonNetwork(params)
# 2. Set continuous inputs (trinary)
network.set_input_states([1, -1, 0, 1, -1])
# 3. Simulate (continuous time)
for step in range(200):
network.simulate_step()
if step % 50 == 0:
outs = network.get_output_states()
print(f"Step {step:3d} | Outputs: {outs} | Energy: {network.get_energy():.3f}")
# 4. Live neuromodulation
network.modulate('dopamine', 0.85) # boost learning
network.modulate('serotonin', 0.6) # modulate plasticity
# 5. Application layer – pattern storage & recall
app = NeuraxonApplication(params)
app.store_pattern("A", [1,1,-1,-1,1], steps=50)
recall = app.recall_pattern("A", steps=30, mask_fraction=0.4)
print("Recall:", recall)
# 6. Evolutionary Aigarth hybrid
hybrid = NeuraxonAigarthHybrid(params)
dataset = [([1,0,0,0,0],[1,0,0,0,0]), ...] # your task
hybrid.evolve(dataset, seasons=5, episodes=20)
print("Best fitness:", hybrid.best().fitness)
# 7. Save / load (full state, backward-compatible with v1)
from neuraxon2 import save_network, load_network
save_network(network, "my_net_v2.json")
loaded = load_network("my_net_v2.json")If you use Neuraxon v 2.0 in your research, please cite:
@article{Vivancos-Sanchez-2026neuraxon2,
title={Neuraxon v2.0: A New Neural Growth \& Computation Blueprint},
author={David Vivancos and Jose Sanchez},
year={2026},
journal={ResearchGate Preprint},
institution={Artificiology Research, UNIR University, Qubic Science},
url={https://www.researchgate.net/publication/400868863_Neuraxon_V20_A_New_Neural_Growth_Computation_Blueprint}
}If you use Neuraxon 1.0 in your research, please cite:
@article{Vivancos-Sanchez-2025neuraxon,
title={Neuraxon: A New Neural Growth \& Computation Blueprint},
author={David Vivancos and Jose Sanchez},
year={2025},
journal={ResearchGate Preprint},
institution={Artificiology Research, UNIR University, Qubic Science},
url={https://www.researchgate.net/publication/397331336_Neuraxon}
}If you use the Neuraxon2LifeTS dataset, please also cite:
@dataset{Neuraxon2LifeTS,
title={Neuraxon 2.0 Time Series Dataset},
author={Vivancos, David and Sanchez, Jose},
year={2026},
publisher={Hugging Face},
url={https://huggingface.co/datasets/DavidVivancos/Neuraxon2LifeTS}
}If you use the NeuraxonLife2-1M dataset, please also cite:
@dataset{NeuraxonLife2-1M,
title={Neuraxon: Artificial Life 2.0 BioInspired Neural Network Simulation 1M Dataset},
author={Vivancos, David and Sanchez, Jose},
year={2025},
publisher={Hugging Face},
url={https://huggingface.co/datasets/DavidVivancos/NeuraxonLife2-1M}
}If you use the NeuraxonLife2.5-100K-TimeSeries dataset, please also cite:
@dataset{NeuraxonLife2.5-100K-TimeSeries,
title={Neuraxon: Artificial Life 2.5 BioInspired Neural Network Simulation 100K-TimeSeries Dataset},
author={Vivancos, David and Sanchez, Jose},
year={2025},
publisher={Hugging Face},
url={https://huggingface.co/datasets/DavidVivancos/NeuraxonLife2.5-100K-TimeSeries}
}We welcome contributions! Areas of interest include:
- Novel plasticity mechanisms
- Additional neuromodulator systems
- Energy efficiency optimizations
- New application domains
- Visualization tools
- Performance benchmarks
- Game of Life extensions and scenarios
Please open an issue to discuss major changes before submitting PRs.
David Vivancos
Artificiology Research https://artificiology.com/ , Qubic https://qubic.org/ Science Advisor
Email: vivancos@vivancos.com
Jose Sanchez
UNIR University, Qubic https://qubic.org/ Science Advisor
Email: jose.sanchezgarcia@unir.net
MIT License. See LICENSE file for details.
Core Neuraxon: Licensed under MIT License (permissive, no restrictions)
Aigarth Hybrid Features: If you implement the Aigarth hybrid features described in our paper, you MUST comply with the Aigarth License, which includes:
- ❌ NO military use of any kind
- ❌ NO use by military-affiliated entities
- ❌ NO dual-use applications with military potential
The standalone Neuraxon implementation (without Aigarth integration) has no such restrictions.
This work builds upon decades of neuroscience research on:
- Synaptic plasticity (Bi & Poo, 1998)
- Neuromodulation (Brzosko et al., 2019)
- Spontaneous neural activity (Northoff, 2018)
- Continuous-time neural computation (Gerstner et al., 2014)
Special thanks to the Qubic's Aigarth team for the evolutionary tissue framework integration.