# From Worms to Mice: Homeostasis Maybe All You Need

**Authors:** Jesus Marco de Lucas  
**arXiv:** [2412.20090](https://arxiv.org/abs/2412.20090)  
**Date:** 2024-12-28  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2412.20090v1/

## Key points

- Author Jesus Marco de Lucas explores neural networks in machine learning, focusing on a simple neural XOR motif with excitatory and inhibitory connections.
- The XOR motif serves as a fundamental mechanism for plasticity in neural circuits, identifying discrepancies between incoming and reference signals while regulating homeostasis.
- The core XOR motif operates on a 4:1 ratio of excitatory to inhibitory neurons and facilitates broader patterns like the 'winner takes all' mechanism.
- Published connectomes across various organisms show the prevalence of the XOR motif at different scales, from tens to millions of neurons.
- Biological neural plasticity is primarily steered by a rudimentary control system that has adapted throughout evolution despite increasing complexity.
- A comparison is made between "strict" XOR motifs and "virtual true" ones with additional connections among nodes, particularly involving specific GABA inhibitory interneurons.
- The commentary highlights the implications of neural XOR motifs in understanding plasticity within neural circuits across diverse organisms, emphasizing the role of homeostasis and basic control mechanisms amidst evolutionary complexities.

## Summary

In his commentary titled "From Worms to Mice: Homeostasis Maybe All You Need," author Jesus Marco de Lucas delves into the fascinating realm of neural networks in machine learning. He explores the central premise of a simple neural XOR motif that incorporates both excitatory and inhibitory connections, positing it as a fundamental mechanism for plasticity within neural circuits of living organisms. This motif serves as a beacon for identifying discrepancies between incoming signals and reference signals, establishing a foundational basis for a loss function in learning neural circuits while concurrently regulating homeostasis. The core XOR motif operates on a 4:1 ratio of excitatory to inhibitory neurons and facilitates broader neural patterns such as the renowned 'winner takes all' (WTA) mechanism. Through an exploration of published connectomes across various organisms with escalating complexity, ranging from C. elegans to Drosophila neuropils and even the mouse V1 visual cortex, the prevalence of the XOR motif is observed at varying scales - from tens to millions and beyond tens of millions. By drawing parallels to machine learning models, de Lucas proposes that biological neural plasticity is primarily steered by a rudimentary control or regulatory system that has endured and adapted throughout evolution despite increasing organism complexity. Furthermore, de Lucas makes an intriguing comparison between a basic "strict" XOR motif and an illustrative example of a "virtual" one featuring additional connections among nodes. The analysis reveals a substantial number of these "virtual true" XOR motifs in connectomes, predominantly associated with specific GABA inhibitory interneurons like AVJR, AVJL, RIBR/L, and RIS. In essence, this commentary sheds light on the profound implications of neural XOR motifs in understanding plasticity within neural circuits across diverse organisms. It underscores the significance of homeostasis as a guiding principle in shaping biological neural plasticity while highlighting the enduring relevance of basic control mechanisms amidst evolutionary complexities.