Coua LabsGet in touch

Research

Following
the questions.

Our interests span learning, reasoning, adaptation, and computational structure. Each project is a way to examine a specific part of that wider landscape.

Paths of inquiryA shared bundle of fine threads branches into three distinct paths, suggesting several directions of investigation.
Paths of inquiryConceptual illustration

First investigation

In preparation

Connectome Learning Priors

Can biological connectivity provide a useful starting structure for learning? Our first project examines this question through controlled computational experiments.

What does the wiring contribute?

A connectome describes the connections in a nervous system. These wiring patterns offer a starting point for asking how the organization of a network affects what it can learn.

We are interested in situations where learning signals are limited. The experiments will examine how a network’s initial structure and learning rules interact under those constraints.

Making a fair comparison.

A biological wiring pattern may differ from another network in many ways. To understand what its organization contributes, we need controls that account for simpler differences in network size, connection counts, and training conditions.

Planned network comparisons
ComparisonWhat it helps us examine
Random networksMatch the number of connections to examine what organization adds.
Rewired networksPreserve each node’s connection counts while changing the wiring pattern.
Synthetic structuresVary patterns of local and long-range connectivity.
Conventional architecturesTrain computational networks under comparable conditions.

Preparing the ground.

We are developing the experiments and tools for this work. As the research develops, we aim to share methods, code, and findings that others can inspect and reproduce.

Follow the code on GitHub How we work