This study shows that most variability in neuronal activity across various species and brain regions can be explained by direct dependencies on individual inputs rather than complex interactions between them. Using maximum entropy models equivalent to logistic artificial neurons, researchers found these minimal models capture over 90% of a neuron's variability in the mouse hippocampus and visual cortex, as well as significant portions in C. elegans. The research demonstrates that higher-order correlations and time-delayed dependencies are largely predictable from these simple instantaneous inputs, suggesting most neurons function similarly to perceptrons.
- Direct dependencies explain the vast majority of neuronal activity variability across species.
- Minimal models are mathematically equivalent to logistic artificial neurons or perceptrons.
- Inferred network weights exhibit biological features such as sparsity, heavy-tailed distributions, and directedness.
- Neural communication is highly redundant and remains robust even after significant input loss.
- Direct dependencies explain the vast majority of neuronal activity variability across species.
- Minimal models are mathematically equivalent to logistic artificial neurons or perceptrons.
- Inferred network weights exhibit biological features such as sparsity, heavy-tailed distributions, and directedness.
- Neural communication is highly redundant and remains robust even after significant input loss.
