Seeing Is Believing: Brain-Inspired Modular Training for Mechanistic Interpretability

Seeing Is Believing: Brain-Inspired Modular Training for Mechanistic Interpretability

We introduce Brain-Inspired Modular Training (BIMT), a method for making neural networks more modular and interpretable. Inspired by brains, BIMT embeds neurons in a geometric space and augments the loss function with a cost proportional to the length of each neuron connection. This is inspired by t...

Full description

Bibliographic Details
Main Authors: Ziming Liu, Eric Gan, Max Tegmark
Format: Article
Language:English
Published: MDPI AG 2023-12-01
Series:Entropy
Subjects:
brain-inspired artificial intelligence
mechanistic interpretability
modularity
Online Access:https://www.mdpi.com/1099-4300/26/1/41

Similar Items

Similar Items