Our Generative Cooperative Network (GCN) is a deep neural embedding framework which combines multiple neural architectures to create embedding vectors representing aspects of each Rejuve member’s body and health. It is a system of multiple coevolving cooperating neural net agents, as opposed to GAN (Generative Adversarial Network), which is system of two coevolving competing neural agents. This system combines causal models, some of which are Bayesian models such as our Expert Bayesian Net or the Bayesian Neural Network, the Variational Autoencoder (VAE). Correlational models can be folded in as well to support the causal models. It is also able to connect disparate models and disparate data modalities like imaging and blood test results. These combinations form consonant theory sets in a decentralized manner, so that individual theories help each other to predict and model complex systems such as the human body.
This component is connected to OpenCog through a biomedical-AI customized deployment of transformer networks for graphs, which map nodes into embedding vectors that can be processed by deep neural networks. Other models of the embedding framework can be contributed by member researchers and clinics as well as others. These can compose automatically into consonant ensembles of models that improve each other, using automated model-building algorithms on the SingularityNET platform.