20,000 Neurons: Why Adding More Brain Cells Beats A Data Ceilings Every Time

We all know that feeling when a recipe finally clicks. It is satisfying when the components work together perfectly. Your brain is actually doing the same thing, just with billions of neurons. A single neuron responds unreliably. Even when the same stimulus is shown repeatedly, its activity varies from trial to trial. This is just biology. However, neurons compensate for this variability by representing the same information across many neurons. This is known as population coding. The problem arises when many neuronal signals rise and fall together. This shared variability, known as noise correlation, can overlap with the activity pattern that carries information. This possibility puzzled an international team of researchers. They include experts from Kyoto University, Harvard University, and the University of California, Los Angeles. The paper is published in the journal Science Advances. You can read the full details about this work here.

The team faced a fundamental question. Why does the brain have so many neurons if shared fluctuations impose a ceiling on information? The researchers re-analyzed recordings of approximately 18,000 to 21,000 neurons from each mouse's primary visual cortex. They examined how much information about subtle visual stimuli differences could be read from increasingly large groups of neurons. By repeatedly drawing random subpopulations of different sizes, the team found two power laws. These laws retained the same form after adjusting for population size. The first describes the distribution of noise strengths. The second describes how each noise pattern aligns with the signal.

This is good news for anyone thinking about learning. These scaling laws, combined with the constraints of subsampling, allowed the team to predict information growth beyond the observed population sizes. In all five mice, the measured power-law exponents predicted that noise correlations would not place an upper limit on information. Although stronger noise components tended to align more closely with the signal, the signal also extended across less variable activity patterns. This indicates that shared noise slows the growth of information as more neurons are added, but does not bring it to a halt. For three decades, shared neural fluctuations were widely expected to make information saturate. This is what S. Amin Moosavi of UCLA noted. His team’s results show that this is not inevitable.