DavidVivancos/NeuraxonLife2.5-100K-TimeSeries

---
license: cc-by-sa-4.0
tags:
- time-series
- artificial-life
- neural-networks
- evolutionary-algorithms
- reinforcement-learning
- biology
- simulation
- neuraxon
- Qubic
- Aigarth
pretty_name: Neuraxon Life 2.5 Deep Time Series
size_categories:
- 100M<n<1B
---

# NeuraxonLife2.5-100K-DeepTimeSeries: Artificial Life Neuraxon Neural Network Simulation Deep Time Series Dataset

## Dataset Description

The **NeuraxonLife 2.5 Deep Time Series Dataset** is a massive, comprehensive collection of simulation data from the Neuraxon Game of Life environment. It tracks the evolution of over **100,000 autonomous agents ("NxErs")** evolving biologically-plausible neural networks under survival pressures.

This dataset represents a significant expansion over previous versions, featuring "Deep Time Series" exploration with over **279 million plasticity events** and high-resolution **per-agent time series** data. It is designed for validating the Neuraxon paper: *'A New Neural Growth & Computation Blueprint'* by David Vivancos https://www.vivancos.com/ & Dr. Jose Sanchez https://josesanchezgarcia.com/ for Qubic Science. https://qubic.org/

**Paper Reference:** [Neuraxon: A New Neural Growth & Computation Blueprint](https://www.researchgate.net/publication/397331336_Neuraxon)

## Dataset Summary

This dataset provides granular insights into emergent neural computation in an artificial life setting. It covers 2,791 distinct simulation games involving complex neural phenomena.

**Key Features:**
* **Massive Scale:** Data from 2,791 games, tracking 100,828 unique agents (NxErs), over 10 independent Sessions with about 270+ Rounds each.
* **Deep Plasticity:** Captures ~279 million synaptic plasticity events (LTP/LTD) and ~278 million associativity events.
* **Multi-Scale Time Series:**
* **Global:** Network-wide metrics per tick.
* **Per-NxEr:** Individual agent metrics (energy, neurotransmitters, branching ratios) tracked over time (162 million rows).
* **Evolutionary Lineage:** New `clan_history` and `hall_of_fame` tables tracking lineage merging and ancestral success.
* **Neuromodulation:** Detailed tracking of Dopamine, Serotonin, Acetylcholine, and Norepinephrine dynamics and threshold modulations.

**Processing Stats:**
* **Games Processed:** 2,791
* **Total Neurons:** ~3.76 Million
* **Total Synapses:** ~17 Million
* **Total Event Rows:** >1 Billion combined events

## Supported Tasks

* **Neural Dynamics Analysis:** Study phase coherence, branching ratios, and criticality over millions of ticks.
* **Plasticity & Learning:** Analyze STDP, cooperative associativity, and synaptic weight evolution across fast, slow, and meta timescales.
* **Evolutionary Biology:** Track clan formation, lineage survival, and the correlation between neural architecture and fitness.
* **Neuromodulation Research:** Investigate how neurotransmitter gradients influence global network states and agent behavior.
* **Behavioral Correlation:** Link input/output patterns (vision/movement) to internal neural states and survival outcomes.

## Dataset Structure

The dataset consists of 28 interconnected Parquet tables.

```text
/
├── games.parquet # Global game metadata and summary stats
├── nxers.parquet # Agent config, stats, and sensory setup
├── network_params.parquet # Hyperparameters and architecture per agent
├── neurons.parquet # Neuron snapshots (membrane potential, health)
├── synapses.parquet # Synaptic weights, types, and integrity
├── time_series.parquet # Global (Game-level) time series metrics
├── per_nxer_time_series.parquet # Individual (Agent-level) time series metrics (HUGE)
├── plasticity_events.parquet # LTP/LTD specific events
├── weight_evolution_events.parquet # Weight deltas across timescales
├── associativity_events.parquet # Cooperative LTP events
├── neuromodulator_events.parquet # Threshold crossing events
├── threshold_modulation_events.parquet # ACh/Autoreceptor threshold impacts
├── subthreshold_events.parquet # Near-threshold integration data
├── autoreceptor_events.parquet # Autoreceptor specific data
├── homeostatic_events.parquet # Homeostatic threshold adjustments
├── adaptive_threshold_events.parquet # Criticality adjustments
├── dendritic_events.parquet # Dendritic spikes and plateau potentials
├── phase_events.parquet # Phase coherence and reset events
├── spontaneous_events.parquet # Spontaneous firing logs
├── silent_synapse_events.parquet # Synapse activation/deactivation
├── nxer_events.parquet # Life events (Birth, Death, Mating)
├── clan_history.parquet # Lineage and clan merging history
├── hall_of_fame.parquet # Top ranked agents
├── io_patterns.parquet # Input/Output vectors
├── itu_fitness.parquet # ITU/Aigarth fitness tracking
├── synapse_neighbor_ids.parquet # Synaptic topology
├── neuron_state_history.parquet # Recent state buffers
├── nxer_visited_history.parquet # Spatial exploration history
├── foods.parquet # Food source locations
├── food_progress.parquet # Harvest trackers
├── world_grids.parquet # Terrain maps
└── manifest.json # Dataset metadata
```

## Data Tables Detail

### 1. Global & Agent Metadata

* **games.parquet** (2,791 rows)
* *Description:* High-level summary of every simulation run.
* *Key Columns:* `game_id`, `round_number`, `total_ticks`, `peak_network_activity`, `average_branching_ratio`, `peak_dopamine`, `total_plasticity_events`, `clan_count`.
* **nxers.parquet** (100,828 rows)
* *Description:* Static data and final stats for every agent (NxEr) born.
* *Key Columns:* `nxer_id`, `clan_id`, `generation`, `ancestors_count`, `stats_fitness_score`, `stats_temporal_sync_score`, `born_ts`, `died_ts`.
* **network_params.parquet** (100,828 rows)
* *Description:* The genomic/hyperparameter configuration for each agent's brain.
* *Key Columns:* `num_hidden_neurons`, `connection_probability`, `learning_rate`, `stdp_window`, `dopamine_baseline`, `oscialltor_frequencies`, `plasticity_energy_cost`.
* **hall_of_fame.parquet** (83,609 rows)
* *Description:* Rankings of the top agents by various categories (Survival, Exploration, Efficiency).
* *Key Columns:* `category`, `rank`, `name`, `ancestors_json`, `stats_fitness_score`.
* **clan_history.parquet** (14,085,063 rows)
* *Description:* Detailed tracking of clan formations, mergers, and active members.
* *Key Columns:* `clan_id`, `members_count`, `merged_from_json`, `created_at_round`.

### 2. Time Series Data (Temporal)

* **time_series.parquet** (8.5 Million rows)
* *Description:* **Global** simulation metrics recorded every tick.
* *Key Columns:* `tick`, `network_activity`, `branching_ratio`, `phase_coherence`, `total_energy`, `dopamine`, `serotonin`, `acetylcholine`, `norepinephrine`, `itu_mean_fitness`.
* **per_nxer_time_series.parquet** (162.9 Million rows | ~20.9 GB)
* *Description:* **Agent-level** metrics recorded every tick. This is the largest table, allowing deep analysis of individual lifespans.
* *Key Columns:* `nxer_id`, `tick`, `alive`, `network_activity`, `total_energy`, `dopamine`, `mean_w_fast`, `mean_w_slow`, `fitness_score`.

### 3. Neural Structure (Snapshots)

* **neurons.parquet** (3.76 Million rows)
* *Description:* State snapshots of neurons (potential, health, adaptation).
* **synapses.parquet** (17.0 Million rows)
* *Description:* Synaptic connections, including weights (`w_fast`, `w_slow`, `w_meta`) and integrity.
* **world_grids.parquet** (2,791 rows)
* *Description:* The 2D terrain map for each game.

### 4. Event Logs (High Frequency)

* **plasticity_events.parquet** (278.9 Million rows)
* *Description:* Logs every LTP/LTD event.
* *Columns:* `pre_id`, `post_id`, `delta_w`, `type`.
* **associativity_events.parquet** (278.9 Million rows)
* *Description:* Cooperative learning events where neighbor synapses assisted potentiation.
* *Columns:* `own_delta_w`, `neighbor_contribution`, `amplification_factor`.
* **weight_evolution_events.parquet** (279.1 Million rows)
* *Description:* Tracking weight shifts across the three timescales (Fast, Slow, Meta).
* **neuromodulator_events.parquet** (272.0 Million rows)
* *Description:* Events where global modulators crossed affinity thresholds.
* **subthreshold_events.parquet** (277.2 Million rows)
* *Description:* Membrane potential integration dynamics before firing.
* **threshold_modulation_events.parquet** (262.4 Million rows)
* *Description:* Adjustments to firing thresholds via Acetylcholine or Autoreceptors.

## Paper Section Mapping

The dataset is structured to support specific sections of the Neuraxon research paper. Use these field mappings for validation:

| Paper Section | Relevant Columns / Tables |
| :--- | :--- |
| **1. Trinary Neuromodulation** | `dopamine`, `serotonin`, `acetylcholine`, `excitatory_fraction`, `inhibitory_fraction`, `neutral_fraction`, `threshold_modulation_by_ach` |
| **2. Temporal Dynamics** | `oscillator_low/mid/high`, `temporal_sync`, `mean_phase_velocity`, `membrane_potential_mean` |
| **3. Synaptic Computation** | `mean_w_fast`, `mean_w_slow`, `mean_w_meta`, `silent_synapse_count`, `ionotropic_contribution_mean` |
| **4. Plasticity & Adaptation** | `ltp_rate`, `ltd_rate`, `associativity_event_count`, `mean_learning_rate_mod`, `plasticity_events` (table) |
| **5. Complex Signaling** | `dendritic_spike_count`, `mean_plateau_potential`, `subthreshold_integration_count` |
| **6. Self-Generated Activity** | `spontaneous_firing_count`, `mean_autocorrelation_window`, `mean_intrinsic_timescale` |
| **7. Synchronization** | `phase_coherence`, `branching_ratio`, `cfc_low_mid`, `pac_theta_gamma` |
| **8. Aigarth/ITU Evolution** | `itu_mean_fitness`, `itu_mutation_events`, `itu_pruning_events`, `clan_history` (table) |

## Usage

### Loading with Hugging Face Datasets

```python
from datasets import load_dataset

# Load specific tables to save memory (e.g., just the game metadata and global time series)
dataset = load_dataset(
"DavidVivancos/NeuraxonLife2.5-100K-TimeSeries",
data_files={
"games": "games.parquet",
"time_series": "time_series.parquet"
}
)

print(dataset['games'][0])
```

### Loading with Pandas

```python
import pandas as pd

# Load the global time series
df_ts = pd.read_parquet("time_series.parquet")

# Plot global network activity over time for a specific game
game_id = df_ts['game_id'].iloc[0]
subset = df_ts[df_ts['game_id'] == game_id]
subset.plot(x='tick', y='network_activity')
```

### Note on Large Tables
The `per_nxer_time_series.parquet` (20GB+), `associativity_events.parquet`, and `plasticity_events.parquet` tables are very large. It is recommended to load these using streaming or by filtering for specific `game_id`s or `round_number`s if not using a distributed framework like Spark or Dask.

## Citation

```bibtex
@dataset{NeuraxonLife2.5-TimeSeries,
title={Neuraxon Game of Life 2.5 Research Dataset: Deep Time Series Exploration},
author={Vivancos, David and Sanchez, Jose},
year={2026},
publisher={Hugging Face},
version={2.5.0},
url={https://huggingface.co/datasets/DavidVivancos/NeuraxonLife2.5-100K-TimeSeries}
}
```

## Authors & Curators

* **David Vivancos** / Artificiology Research - [Qubic Science](https://qubic.org/)
* **Dr. Jose Sanchez** / UNIR - [Qubic Science](https://qubic.org/)

**Contact:** For questions or issues, please open a GitHub issue at [https://github.com/DavidVivancos/Neuraxon](https://github.com/DavidVivancos/Neuraxon).