Finding Emergence in Data Using Machine Learning
Author Information
Author(s): Yang Mingzhe, Wang Zhipeng, Liu Kaiwei, Rong Yingqi, Yuan Bing, Zhang Jiang
Primary Institution: Beijing Normal University
Hypothesis
Can a machine learning framework effectively quantify causal emergence and model macro-dynamics in complex systems?
Conclusion
The proposed machine learning framework successfully quantifies causal emergence and models macro-dynamics, demonstrating improved generalization across diverse conditions.
Supporting Evidence
- The framework effectively captures emergent dynamics from both simulated and real data.
- NIS+ shows superior generalization capabilities compared to alternative approaches.
- Experiments demonstrate the ability to quantify causal emergence under various conditions.
- NIS+ outperforms other models in multi-step predictions and pattern capturing.
- Real fMRI data analysis indicates that NIS+ can learn and simulate complex brain dynamics.
Takeaway
This study shows how a computer program can learn to find patterns in complex data, like how a group of birds moves together or how our brains react to movies.
Methodology
The study introduces a machine learning framework called NIS+ that maximizes effective information to learn macro-dynamics and quantify causal emergence from data.
Potential Biases
Potential overfitting when using complex models with many parameters.
Limitations
The framework requires a large amount of training data, which may not be feasible in many real-world cases, and the interpretability of the neural networks remains a challenge.
Participant Demographics
830 subjects from fMRI studies, including various demographics watching the same movie clip.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website