The Neuron Is the Distribution

Research overview

This research article develops the architectural thesis that a neuron can be treated as a distribution rather than only as a point-valued activation. It positions variational distributional neurons as local computational units whose posterior state, uncertainty and sampled latent activity participate in the forward computation. The article distinguishes distributional computation from uncertainty placed only on parameters, predictive outputs, ensembles or global latent variables. It also clarifies the relationship between local variational computation, stochasticity and Bayesian modeling, framing EVE as a change in computational granularity rather than as a synonym for a Bayesian neural network.

This JWAYU overview summarizes the article’s research positioning; the publisher page remains the reference for the official abstract and manuscript.

Keywords

• the neuron is the distribution
• distributional computation
• variational neuron
• local posterior
• uncertainty
• computational granularity
• EVE

Citation

Ruffenach, Yves. “The Neuron Is the Distribution.” Research Square, July 14, 2026. Research Article. DOI: 10.21203/rs.3.rs-10333402/v1.

Related research

• Variational Distributional Neuron
• Distributional Neurons: Making Uncertainty a Unit of Computation
• The Neuron as a Latent State: Classical Variational Readout in Distributional Neural Units
• Measuring and Controlling Internal Activity in Variational Neural Units
• Variational Neurons in Transformers for Language Modeling

Variational neuron · Variational Transformer · Internal uncertainty · Variational closure