此页面由机器翻译成中文。 查看英文原版。

Interactive learning

Simulations

12 interactive, browser-based simulations of Active Inference and the Free Energy Principle — from particles self-organising into life to agents navigating a T-maze. Each one runs entirely in your browser. Explore them by difficulty, or dive straight in.

入门级

初学者

从这里开始。直观的视觉入口点进入主动推断和自由能量原则。

Life Emerges — Living Ink

Watch particles self-organise from chaos into a living system by toggling between entropy and free-energy minimization.

Apple or Frog? — Perception as Inference

See how the brain resolves ambiguous sensory data through Bayesian inference by adjusting beliefs and likelihoods.

Surprise Minimisation — The Fish Predicts

Control the environment while a fish minimizes surprise, demonstrating prediction error and belief updating in real time.

Collective Active Inference — Fish Schooling

Multiple fish self-organize into group-level agents, demonstrating Markov blankets and collective behavior.

中级水平

中间级

通过完整的反推循环,观察自我组织在代理人和结构中展开。

From Beliefs to Action — The Complete Loop

An eight-step guided tutorial through the complete Active Inference cycle, from generative models to policy selection.

The Self-Evidencing Tree — Morphogenesis as Inference

A tree that grows by minimizing free energy, exploring the explore–exploit trade-off through structural adaptation.

The Colony as Inference — Active InferAnts

An ant colony where individuals minimize expected free energy through pheromone communication and niche construction.

Active Inference Network — Collective Autonomy

Scale a complete graph from one agent to 150 and watch Markov blankets and collective autonomy emerge.

高级水平

高级

包括精度、预期自由能和嵌套被套在内的规范层级演示。

T-Maze — Expected Free Energy

The canonical Active Inference demonstration: an agent navigates by balancing information-seeking against reward.

Hierarchical Markov Blankets

Visualize nested Markov blankets from cells to organisms, showing recursive self-organization.

The Dopamine Dial — Neuromodulation as Precision

Adjust four neurotransmitter channels in a harvester ant brain to demonstrate precision-weighted behavior control.

Tree Morphology Quiz — Precision, Preferences & Hierarchy

An interactive quiz on how branching growth expresses precision, prior preferences, and hierarchical inference.

关于这些模拟

学习实践中得成(或「以行学」)

These demonstrations make abstract Active Inference concepts tangible. They are also published on the Institute's main site at activeinference.org/pages/simulations.html. To learn the theory behind them, see Active Inference and the Learning and Research resources.