Open source project

ACLAS Neuro-Edu SDK

Explore Neuro-Edu, an ACLAS open-source research framework for AI-powered education and learning experiments.

What matters

ACLAS Neuro-Edu SDK

Neuro-Edu is a research-grade, open-source multi-agent cognitive simulation framework modelling knowledge acquisition.
Its approach combines on-device neural networks, vector-space semantics and thermodynamic entropy analysis, with real backpropagation.
The on-device neural kernel is a pure NumPy MLP with He initialisation, ReLU/Sigmoid activation and stochastic gradient descent; 3-layer architecture 6→16→8→1.
Multi-agent social learning uses 20+ autonomous agents exchanging knowledge via a real-time message bus.
Cognitive entropy analytics include Shannon entropy, MSE loss curves, GPA, CAS scores, retention rates, Shannon Diversity Index and dropout risk prediction.
The knowledge graph is a D3.js concept relation map; a 5-dimension skill matrix tracks Logic, Math, Language, Memory and Creative abilities.
Four-layer architecture: WebGL/3D Nebula; Neural Kernel; Cognitive Profiles; Core Engine + API exposed via FastAPI REST.
REST endpoints: /api/status, /api/teach, /api/train, /api/metrics, /api/graph, /api/architecture, /api/reset.
Pluggable LLM engines: Ollama or vLLM, supporting Llama-3, Qwen 2.5, DeepSeek-V3, Gemma-2, Phi-3, Mistral/Mixtral.
The project includes a CLI, a 50-sample labelled training dataset and 26 passing tests with CI/CD.
The interactive browser demo runs 20 autonomous AI agents with a WebGL neural nebula and real-time telemetry.

Details

Key information

DOI: 10.5281/zenodo.19743807
License: MIT — free for academic & commercial use
Repository: https://github.com/aclascollege/neuro-edu
Model hub: https://huggingface.co/ACLASCollege/neuro-edu-core-v3
Neural architecture: 6→16→8→1; Dense(16)+ReLU → Dense(8)+ReLU → Dense(1)+Sigmoid
Agents: 20+ autonomous agents (demo states 20)
Tests: 26 passing tests (0.52s)
Training dataset: 50-sample labeled training dataset (knowledge_base.json)
Skill matrix: 5 dimensions — Logic, Math, Language, Memory, Creative
Stack: Python 3.10+, FastAPI, NumPy, Three.js, Chart.js, D3.js
Pluggable engines: Ollama, vLLM; Llama-3, Qwen 2.5, DeepSeek-V3, Gemma-2, Phi-3, Mistral/Mixtral
REST endpoints: /api/status, /api/teach, /api/train, /api/metrics, /api/graph, /api/architecture, /api/reset
Citation: ACLAS Neuro-Edu: Autonomous Cognitive Simulation Framework, Atlanta College of Liberal Arts and Sciences, 2026

Important information

Before you enroll

These statements explain the recognition, regulatory and legal conditions that apply.

We don't wrap GPT. We model cognition from first principles.

A research-grade, open-source multi-agent cognitive simulation framework that models human knowledge acquisition using on-device neural networks, vector-space semantics, and thermodynamic entropy analysis. Built with real backpropagation — no API wrappers, just first-principles AI.

Continue exploring

Ready to start?

Applications are open continuously. There are no term dates, no entrance examinations and no campus visits — you can begin as soon as your enrollment is confirmed.