Data Genes
Research company
Data Genes designs neural networks around the structure of the tesseract and studies systems that improve their own ability to learn. Our work is for the people who engineer medicines, read genomes and care for patients.
Focus
A standard neural network starts with no notion of shape or symmetry. We build geometry into the architecture. The organising structure is the tesseract, the four-dimensional cube: 16 corners, 32 edges, 24 faces and 8 cells, each with a defined relationship to the others.
Where that structure helps is the question we work on. We test it. We do not assume it.
A system that improves with experience is familiar. A system that improves its own ability to improve is a harder problem, and it is the centre of our work. We study it in populations of software agents that inherit traits, vary and face selection, the same logic that shapes living genomes.
We treat it as a question of measurement. The test is fixed before the experiment runs.
Fields
We are at the research stage, and we are looking for a small number of design partners in three fields.
Teams who discover, design and manufacture medicines.
Researchers who work with long biological sequences, where structure appears at many scales at once.
Physicians and clinical researchers who hold a model to the standard of evidence they apply to any other tool.
How we work
We decide how a result will be judged before we run the experiment. We keep the results that improve learning.
Contact
If your team works in one of these fields and wants to explore a research partnership, write to us. Tell us what you work on and what you would want a model to do.