NeuronSoup: Evolving Asynchronous, Shared-Neuron Temporal Graphs without Backpropagation
2026-07-16 • Neural and Evolutionary Computing
Neural and Evolutionary ComputingMachine Learning
AI summaryⓘ
The authors introduce NeuronSoup, a new kind of neural network that processes signals asynchronously through shared neurons rather than in fixed layers. In this system, signals travel through variable routes where neurons handle multiple paths, causing interactions based on timing and signal polarity. They use a genetic algorithm to evolve the network's structure and parameters, achieving decent accuracy on a digit classification task without relying on traditional backpropagation. This approach allows the network to adjust its computation dynamically and discover complex interactions automatically.
asynchronous processingneural networksgenetic algorithmssignal propagationMNIST datasetResNet18constructive interferencedestructive interferenceco-evolutioncomputational topology
Authors
Subodh Kalia
Abstract
We present NeuronSoup, a neural computation architecture that replaces synchronous layer-by-layer processing with asynchronous, delay-mediated signal propagation through a pool of shared neurons. Each path in the network routes a continuous-valued signal from one input neuron to one output neuron through a variable number of intermediate hidden neurons. Hidden neurons are physically shared across paths: when two paths pass through the same neuron, the second arrival encounters the accumulated state left by the first, producing constructive or destructive interference that depends on signal polarity and arrival timing. The entire architecture -- topology, weights, delays, and connectivity -- is co-evolved by a genetic algorithm operating on a flat real-valued genome of 14,602 genes. On 10-class MNIST digit classification using frozen ResNet18 features as input, the system evolves a network of 204 active paths through 266 hidden neurons (156 shared across multiple paths, with one neuron participating in 11 distinct paths) and achieves 85.9\% test accuracy after 10,000 generations. The trained model occupies 115 KB. We argue that this architecture addresses fundamental limitations of current deep learning: it requires no differentiable computation graph, adapts its computation depth per-sample, and discovers lateral interactions between processing pathways that current architectures must engineer explicitly. We discuss why genetic algorithms are the correct optimization tool for this problem class, why CMA-ES fails at this scale, and how the architecture generalizes to arbitrary domains by substituting the encoder and output structure.